DEPARTMENT OF HEALTH AND HUMAN SERVICES
FOOD AND DRUG ADMINISTRATION
CENTER FOR DRUG EVALUATION AND RESEARCH
JOINT MEETING OF THE
CDER PSYCHOPHARMACOLOGIC DRUGS ADVISORY COMMITTEE
FDA PEDIATRIC ADVISORY COMMITTEE
Holiday Inn Bethsda
PSYCHOPHARMACOLOGIC DRUGS ADVISORY
Wayne K. Goodman, M.D., Chair
Jean E. Bronstein, R.N., M.S.
James J. McGough, M.D.
Philip S. Wang, M.D., M.P.H., Dr.P.H.
Lauren Marangell, M.D.
Dilip J. Mehta, M.D., Ph.D.
Delbert G. Robinson, M.D.
Daniel S. Pine, M.D.
Barbara G. Wells, Pharm.D.
Bruce G. Pollock, M.D., Ph.D.
PEDIATRIC ADVISORY COMMITTEE MEMBERS:
P. Joan Chesney, M.D., Chair
Deborah L. Dokken, M.P.A.
Michael E. Fant, M.D., Ph.D.
Richard L. Gorman, M.D.
Robert M. Nelson, M.D., Ph.D.
Thomas B. Newman, M.D., M.P.H.
Judith R. O'Fallon, Ph.D.
Victor M. Santana, M.D.
CONSULTANTS AND GUESTS (Voting):
Norman Fost, M.D., M.P.H.
Charles E. Irwin, Mr., M.D.
Laurel K. Leslie, M.D., F.A.A.P.
Steven Ebert, Pharm.D. (Consumer Representative)
James M. Perrin, M.D.
Cynthia R. Pfeffer, M.D.
Gail W. Griffith
(Patient Representative, Voting)
Robert D. Gibbons, Ph.D.
Tana A. Grady-Weliky, M.D.
Richard P. Malone, M.D.
Irene E. Ortiz, M.D.
Matthew V. Rudorfer, M.D.
GUEST SPEAKERS AND GUESTS (Non-Voting):
Kelly Posner, Ph.D.
John March, M.D., M.P.H.
Samuel Maldonado, M.D., M.P.H.
Barbara Stanley, Ph.D.
Madelyn Gould, Ph.D., M.P.H.
Robert Temple, M.D.
Russell G. Katz, M.D.
Thomas Laughren, M.D.
M. Dianne Murphy, M.D.
Anne Trontell, M.D., M.P.H
Anuja M. Patel, M.P.H., Executive Secretary
C O N T E N T S
Call to Order and Opening Remarks,
Wayne Goodman, M.D. 6
Introduction of Committee 9
Conflict of Interest Statement,
Anuja Patel, M.P.H. 20
Overview of Issues:
Dianne Murphy, M.D., Director, Office of
Office of the Commissioner 25
Russell Katz, M.D., Director, Division of
Neuropharmacological Drug Products,
DNDP, CDER 34
Regulatory History and Background,
Thomas Laughren, M.D., Team Leader, DNDP, CDER 46
Recent Observational Studies of Antidepressants
and Suicidal Behavior,
Diane Wysowski, Ph.D., Division of Drug Risk
Evaluation, Office of Drug Safety, CDER 62
Brief Report on TADS Trial,
John March, M.D., M.P.H., Duke University 74
Committe Discussion on TADS Trial 93
Characteristics of Pediatric Antidepressant Trials,
Greg Dubitsky, M.D., Medical Officer,
DNDP, CDER 107
Classification of Suicidality Events,
Kelly Posner, Ph.D., Columbia University 117
OCTAP Appraisal of Columbia Classification
Solomon Iyasu, M.D., M.P.H., Team Leader,
Office of Counter-Terrorism and Pediatric
Drug Development 140
C O N T E N T S (Continued)
Results of the Analysis of Suicidality in Pediatric
Trials of Newer Antidepressants,
Tarek Hammad, M.D., Ph.D., M.Sc., M.S., Senior
Medical Reviewer, DNDP, CDER 152
Comparison Between Original ODS and DNDP Analyses
of Pediatric Suicidality Data Sets,
Andrew Mosholder, M.D., M.P.H., Division of
Drug Risk Evaluation, ODS, CDER 200
Citalopram and Escitalopram Product Safety Data,
Jeffrey Jonas, M.D., Forest Laboratories, Inc. 219
Sertraline Use in Pediatric Population: A
Steven J. Romano, M.D., Pfizer, Inc. 232
Wyeth Pharmaceuticals, Joseph S. Camardo, M.D. 247
Open Public Hearing 255
Summary by the Committee Chair 444
P R O C E E D I N G S
Call to Order and Opening Remarks
DR. GOODMAN: I wish to welcome you to
this two-day joint session of the
Psychopharmacologic Drugs Advisory Committee and
the Pediatric Advisory Committee, being held on
September 13th and 14th here, at the Holiday Inn in
I am Wayne Goodman, Professor of
Psychiatry at the University of Florida, today
wearing my hat as chair of the advisory committee.
As you settle in, please take this opportunity to
put into silent mode your cell phones and any other
devices that emit sounds in the audible range of
Some of you may be surprised not to see
Matt Rudorfer in this seat but we arm-wrestled for
the position and he won.
In all seriousness, his term has ended but
we are fortunate to see him return as a voting
consultant to the committee.
I have some official language to read to
you. All committee members and consultants have
been provided with copies of the background
materials, from both the sponsors and the FDA, and
with copies of letters from the public that we
received by the August 23rd deadline. The
background materials have been posted on the FDA
website. Copies of all these materials are
available for viewing at the FDA desk outside this
We have a very large table, a full house
and important topic today so I would like to start
with a few rules of order. Please speak directly
into the mike when called on. We will be keeping
track of individuals at the table who wish to speak
and we will call upon them in order.
FDA relies on the advisory committee to
provide the best possible scientific advice
available to assist us in the discussion of complex
topics. We understand that issues raised during
the meeting may well lead to conversations over
breaks or during lunch. However, one of the
benefits of an advisory committee meeting is that
the discussions take place in an open and public
forum. To that end, we request that members of the
committee not engage in off-record conversations on
today's topic during the breaks and lunch.
Whenever there is an important topic to be
discussed there are a variety of opinions. One of
our goals today and tomorrow is for the meeting to
be conducted in a fair and open way where every
participant is listened to carefully and treated
with dignity, courtesy and respect. Anyone whose
behavior is disruptive to the meeting will be asked
to leave. We are confident that everyone here is
sensitive to these issues so understand that these
comments are as a gentle reminder.
We look forward to a productive and
interesting meeting. This is an unusual meeting in
that we have two advisory committees represented
here, Psychopharmacological Drugs Advisory
Committee, chaired by myself, and the Pediatric
Advisory Committee, chaired by Joan Chesney, to my
left. We will now go around the table and have the
committee introduce themselves, starting on my
right. Please indicate your expertise and
affiliation. We will start in that corner, over
DR. TEMPLE: Bob Temple. I am the Office
Director, ODE I.
DR. KATZ: Russ Katz, Division Director,
Division of Neuropharmacological Drug Products,
DR. LAUGHREN: Tom Laughren, phychopharm.
team leader, in the Neuropharmacological Division.
DR. MURPHY: Dianne Murphy, Office
Director, Office of Pediatric Therapeutics.
DR. TRONTELL: Anne Trontell, Deputy
Director, Office of Drug Safety.
DR. FANT: I am Michael Fant, University
of Texas Health Science Center in Houston. My
expertise is neonatology and biochemistry.
DR. PFEFFER: Cynthia Pfeffer. I am a
child psychiatrist at Weill Medical College of
Cornell University, and I have expertise
depression suicidal behavior in children and
DR. FOST: Norm Fost, University of
Wisconsin, Professor of Pediatrics, Director of the
Bioethics Program and Chair of the IRB.
DR. ORTIZ: Irene Ortiz, University of New
Mexico, Albuquerque VA. My expertise is in
depression in the elderly.
DR. MALONE: Richard Malone, Drexel
University College of Medicine, and my area is
DR. NELSON: Robert Nelson, Children's
Hospital of Philadelphia and the University of
Pennsylvania. My expertise is in pediatric
critical care medicine and ethics.
DR. PERRIN: Jim Perrin, Professor of
Pediatrics, Harvard Medical School and Head of the
Division of General Pediatrics at the Mass. General
Hospital. I have shortened my expertise as being
in general pediatrics.
DR. GRADY-WELIKY: Tana Grady-Weliky,
Associate Professor of Psychiatry at the
of Rochester School of Medicine and Dentistry. My
expertise is in mood disorders and women across the
reproductive life cycle and medical education.
DR. EBERT: Steven Ebert, Department of
Pharmacy of Meriter Hospital and School of
Pharmacy, University of Wisconsin, Madison.
DR. GIBBONS: Robert Gibbons, Professor of
Statistics and Professor of Psychiatry and Director
of the Center for Health Statistics at the
University of Illinois, Chicago. I only do math!
DR. PINE: Danny Pine, child and
adolescent psychiatrist, National Institute of
Mental Health intramural research program. I am a
clinical child psychiatrist.
MS. BRONSTEIN: Jean Bronstein,
psychiatric nurse, Stanford University Hospital,
the consumer representative.
DR. RUDORFER: Matthew Rudorfer, National
Institute of Mental Health. My areas of expertise
are mood disorders and psychopharmacology.
MS. PATEL: Anuja Patel, Advisors and
Consultants Staff, Executive Secretary
Psychopharmacologic Drugs Advisory Committee.
DR. CHESNEY: Joan Chesney, the University
of Tennessee, in Memphis, and Professor of
Pediatrics, and my specialty is infectious
DR. MCGOUGH: Jim McGough, Professor of
Psychiatry, UCLA. My area is child and adolescent
MS. GRIFFITH: My name is Gail Griffith
and I serve as the patient rep. on this committee,
and I would just like to take this opportunity to
say why I am here. First, I am not a medical
professional; I am a consumer. I have suffered
from major depression since I was a teen. Second,
I have a son who suffers from major depression and
three years ago, at age 17, after he was diagnosed
and placed on a regimen of antidepressants he
attempted suicide by overdosing intentionally on
all his medications. He nearly died. So, I know
this illness. I know what it does to adolescents.
For the record, I would simply like to
state that I have no professional ties to
advocacy group or any patient constituency. I also
wish to affirm that I have no ties to any
pharmaceutical company, nor do I hold any
investments in pharmaceutical manufacturers. My
sole responsibility is to ensure that the interests
of concerned parents and families are represented
at this meeting.
DR. MARANGELL: Lauren Marangell, Baylor
College of Medicine. I specialize in adult
interventions in mood disorders, both unipolar and
DR. ROBINSON: I am Delbert Robinson. I
am from the Albert Einstein College of Medicine, in
New York, and I specialize in psychotic disorders
and anxiety disorders.
DR. LESLIE: Laurel Leslie. I am a
behavioral developmental pediatrician at Children's
Hospital, San Diego and my area of expertise is in
children's mental health services research.
DR. IRWIN: Charles Irwin. I am a
professor of pediatrics at the University of
California, San Francisco. I am in charge of the
Division of Adolescent Medicine at the University,
which is a multi-disciplinary program that cares
for adolescents and trains large numbers of
individuals caring for teenagers, and my research
is in the area of risk-taking during adolescence.
MS. DOKKEN: I am Deborah Dokken. I
reside in the Washington, D.C. Metro area. I do
not have a specific institutional affiliation, and
I have for several years been involved in parent
and family advocacy and health care.
DR. NEWMAN: I am Thomas Newman. I am a
professor of epidemiology and biostatistics in
pediatrics at the University of California, San
Francisco, and a general pediatrician.
DR. WELLS: I am Barbara Wells. I am a
professor and Dean of the School of Pharmacy at the
University of Mississippi. My expertise is in
DR. POLLOCK: I am Bruce Pollock. I am a
professor of psychiatry, pharmacology and
pharmaceutical sciences at the University of
Pittsburgh. I head the Division of Geriatric
Psychiatry at the university.
DR. O'FALLON: Judith O'Fallon, Emeritus
Professor of Biostatistics from the Mayo Clinic,
with 30 years of experience particularly in cancer
clinical trials but clinical trials methods.
DR. SANTANA: Good morning. I am Victor
Santana. I am a pediatric hematologist/oncologist
at St. Jude's Children's Research Hospital in
DR. WANG: I am Philip Wang, Harvard
Medical School. I am a psychiatrist and
epidemiologist and those are my areas of expertise.
DR. GORMAN: Richard Gorman, a practicing
pediatrician for 20 years in the Baltimore suburbs,
Chair of the American Academy's Committee on Drugs,
and representing the American Academy of Pediatrics
at this table.
DR. MALDONADO: Sam Maldonado. I work at
pediatric drug development at Johnson & Johnson. I
am one of the industry representatives to this
DR. MEHTA: Dilip Mehta, retired industry
executive and industry representative on the
Psychopharmacologic Drugs Advisory Committee.
DR. GOODMAN: Thank you, all, for being
with us these two days. Our session today is the
second of two planned advisory committee meetings,
convened to address recent concerns about reports
of suicidal ideation and behavior developing in
some children and adolescents during treatment of
depression with a selective serotonin reuptake
inhibitor, an SSRI, or other newer generation
antidepressants. Our goal is to gather information
from a variety of sources and perspectives to help
us understand this complex situation, and
ultimately to offer the best possible
recommendations to the FDA.
I would like to thank the many groups,
individuals and families that submitted written
statements in advance of the meeting, many of which
were quite informative as well as moving. A major
portion of today's meeting will be devoted to a
four-hour open public hearing during which dozens
of people from around, and even beyond,
will have the opportunity to present their own
personal or professional experiences and ideas
about the relative risks and benefits of
antidepressant medication in children and
adolescents. Although the necessary consideration
of the clock will permit only a short time at the
microphone for each speaker, I can assure you that
the committee welcomes and values input from all
viewpoints and feels it is essential to our work
that all voices be heard.
The committee's task is more difficult
than usual. Our review is not confined to whether
one agent is safe and effective based upon the
corresponding clinical trials submitted to the FDA.
We are faced, instead, with assessing efficacy and
safety for nine drugs that represent more than one
chemical class of antidepressants, all of which are
already available on the market.
Although the cornerstone of the data under
examination is derived from randomized clinical
trials submitted to the FDA this time, following a
reclassification of the adverse events,
ourselves turning to information from a wide
variety of sources, in particular to inform
ourselves about the drugs' possible benefits in
this population. However, once we open our minds
to consideration of data originating outside
randomized clinical trials we rest upon a slippery
slope in which variations in interpretation are
introduced according to the weighting each member
places on the merits of the source.
For me, the difficulty in assessing the
balance between benefit and risk is multiplied by
the nature of the adverse events under scrutiny.
Psychiatrists grapple, for the most part, with
illnesses that produce significant morbidity and
more rarely mortality except from suicide. Nothing
in my experience is more tragic than the loss of a
child to suicide. To think that I might prescribe
an agent that contributed to that outcome is
unbearable. Equally unbearable is to think that I
did not do enough to prevent it. This is the
essence of the dilemma before us.
We may not have all the data we
like, especially to assess long-term benefit. We
can make recommendations about what research should
be conducted, but we will be faced at the
conclusion of business tomorrow to make
recommendations based upon what we know at this
cross-section in time. In deliberating on the
safety of antidepressant treatment in children, let
us not forget the toxicity of the underlying
disease. Major depression remains an
under-diagnosed, under-studied and under-treated
serious disorder among many thousands of our
nation's youth, leading to considerable suffering,
disability and heartbreak in many families.
I believe that all of us in this room
share the desire to alleviate the suffering from
this disorder through the successful use of
interventions that are made available to all those
who need them. Despite the daunting task before
us, I remain hopeful that with a fair and
open-minded review of the evidence this advisory
committee will constructively address the issues
and ensure that interventions for this
disorder meet high standards for both effectiveness
Now I will ask Anuja Patel, executive
secretary for the advisory committee, to review
some of the ground rules for this committee and the
Conflict of Interest Statement
MS. PATEL: Good morning. Before I
continue, I would like to notify you of a
correction on the roster attached to the agenda.
The following consultants, Dr. Robert Gibbons, Dr.
Matthew Rudorfer, Dr. Richard Malone, Dr. Tana
Grady-Weliky and Dr. Irene Ortiz will be added to
the roster. Amended copies of the roster will be
available later this morning at the information
desk outside this ballroom.
As you know, we have a very full open
public hearing today, and in the interest of both
fairness and efficiency we are running it by some
strict rules. To make transitions between speakers
more efficient, all speakers will be using the
microphone and podium in front of the
Each speaker has been given their number in the
order of presentations and when the person ahead of
you is speaking, we ask that you move to the nearby
next speaker chair. Individual presenters and
families have been allotted three minutes for their
presentations. The one consolidated presentation
has been given five minutes. We will be using a
timer and speakers who run over their time will
find that the microphone is no longer working. We
apologize for the need for the strict rules, but we
wanted to be fair and to give as many people as
possible an opportunity to participate.
The public may submit comments after this
meeting directly to the FDA's Division of Dockets
Management. Instructions for submitting electronic
and written statements are available at the
registration desk outside this room. The docket
will remain open until July 29, 2005. Thank you
for your cooperation.
I would like to read the meeting statement
into the record now. The following announcement
addresses the issue of conflict of
interest and is
made a part of the record to preclude even the
appearance of such at this meeting. The topics to
be discussed today are issues of broad
applicability. Unlike issues before a committee in
which a particular company's product is discussed,
issues of broader applicability involve many
industrial sponsors and products.
All special government employees and
invited guests have been screened for their
financial interests as they may apply to the
general topics at hand. The Food and Drug
Administration has granted particular matters of
general applicability waivers under 18 USC
208(b)(3) to the following special government
employees, which permits them to participate fully
in today's discussion and vote: Jean Bronstein,
Dr. Joan Chesney, Dr. Wayne Goodman, Dr. Lauren
Marangell, Dr. James McGough, Dr. James Perrin, Dr.
Bruce Pollock. In addition, Dr. Philip Wang has
been granted a limited waiver that permits him to
participate in the committee's discussions. He is,
however, excluded from voting.
A copy of the waiver statements may be
obtained by submitting a written request to the
agency's Freedom of Infection Office, Room 12A-30
of the Parklawn Building.
In addition, Dr. Judith O'Fallon and Dr.
Victor Santana have financial interest under 5 CFR,
Part II, Sec. 40.202 that are covered by a
regulatory waiver under 18 USC 208(b)(2).
Because general topics impact so many
entities, it is not practical to recite all
potential conflicts of interest as they may apply
to each member, consultant and guest speaker. FDA
acknowledges that there may be potential conflicts
of interest but, because of the general nature of
the discussion before the committee, these
potential conflicts are mitigated.
With respect to FDA's invited industry
representatives, we would like to disclose that Dr.
Dilip Mehta and Dr. Samuel Maldonado are
participating in this meeting as industry
representatives, acting on behalf of regulated
Dr. Mehta is retired from Pfizer and Dr.
Maldonado is employed by Johnson & Johnson.
With respect to all other participants, we
ask in the interest of fairness that they address
any current or previous financial involvement with
any firm whose product they may wish to comment
upon. Thank you.
DR. GOODMAN: We will now proceed with a
series of formal presentations that will bring us
to 11:45 a.m. and then a 15-minute discussion
before lunch. In the interest of time, I would
like to ask my fellow committee members to restrict
their questions after each presentation to issues
of clarification only. There will be time, 15
minutes, for some discussion between 11:45 and
12:00 and tomorrow there will be a great deal of
time for discussion and consideration of the
questions before us. So, please, if you have
questions about clarification, you can ask them
after each presentation but restrict it to those
kinds of issues.
With that, I would like to introduce Dr.
Dianne Murphy, of the FDA, who will be
Dr. Russell Katz, also of the FDA.
Overview of Issues
DR. MURPHY: Good morning and welcome to
this very important discussion. Before we begin
today's important deliberations, I would ask us to
step back and see the broader context in which this
meeting is occurring. I am going to spend a few
minutes trying to describe that for you.
There are four points I hope you take from
this short presentation. One is that the majority
of medicines given to children in this country are
prescribed off-label and have not been studied in
all the pediatric populations in which they are
Second, because of new legislation and
regulations since 1998, FDA has seen an increase in
products that are used in children being studied in
Third, for the first 100 products,
involving over 200 studies conducted as a result of
the new legislation, FDA has found that
approximately one-fourth of the time
there was a
need to change the dose, a new pediatric-specific
adverse event was described or the product was not
found to be efficacious despite the fact that it
was efficacious in adults.
Fourth, part of the reason we are here
today is because we are finally studying the
therapies that are being given to children.
Children deserve the same level of evidence that is
required for adults to determine that their use by
them is safe, effective and properly dosed. They
are a heterogeneous group who undergo rapid
metabolic, hormonal, physiologic, development and
growth changes in comparison to us, adults, who are
rather static and tend only to deteriorate.
Over the last two decades FDA has actively
supported, along with the American Academy of
Pediatrics and many other groups, the efforts to
encourage development of information and
appropriate use of therapies in the pediatric
Very quickly, and this is important to
understand, again, the context in which
this information has been brought to you, in the
last decade we have made tremendous progress. In
1994 the agency published an approach that it hoped
would help foster and encourage development of
therapies that we be used in children. Congress
passed legislation in 1997 which is referred to as
the exclusivity or the incentive to develop studies
on products that are being used in children.
In 1998 the FDA published the Pediatric
Rule, which was an effort to say that if a sponsor
is going to develop a product in adults and that
same disease occurs in children, or condition, that
product in most circumstances and certain
conditions would be required to be studied.
We are going to go more into the 2001
adoption by FDA of Subpart D, Pediatric Ethics
Regulations, and I wanted to bring up the Best
Pharmaceuticals for Children Act, which you will
hear referred to as BPCA, because it renewed the
congressional legislation of '97 and is important
in that, again, it is the renewal of the incentive
to study products that are being used in
Another congressional legislative
activity, the Pediatric Research Equity Act, in
essence confirmed FDA's authority to require
studies in children in certain circumstances.
In the last decade, particularly really
since 1997, the FDA has issued over 290 written
requests to sponsors asking them to study products
in children because these products are being used
in children. We have had submitted to us over 110
products, involving over 220 studies in children,
and have now more than 76 new labels that have new
pediatric information from these studies.
The major depressive disorders were
included in the written requests that were issued,
and written requests were issued for the products
you see listed here, Prozac, Zoloft, Remeron, Paxil
Effexor, Celexa and Serzone. Those studies were
all conducted under this program or in response to
This is a list of some of the programs and
activities that are in place at FDA to help ensure
the quality and ethical conduct of
studies and the
approach to pediatrics. This is really focusing
mostly on the drugs component of this. But there
is, at the Commissioner's level, an Office of
Pediatric Therapeutics. This was enacted by the
Best Pharmaceuticals for Children Act in 2002 and
first staff were hired last year. We now have in
place an ethicist whose focus is pediatric ethics.
You will hear a little bit more about the Pediatric
Advisory Committee and Subpart D referrals in a
minute. You just heard about the exclusivity
process which has been important in making sure
that trials do get conducted. I will spend a few
moments at the end talking about disclosure
requirements that are unique to pediatric studies
that are conducted under exclusivity.
This is another meeting, actually in a
long series of meetings that have occurred to
ensure the scientific and ethical quality of
activities involving studies that are being
conducted in children. Since 1999, the Pediatric
Advisory Subcommittee has had, including today's
meeting, over ten meetings that have
ten scientific issues, three ethical issues and, in
addition, starting last year, began having specific
safety reviews of products that have been approved,
again under the exclusivity provisions, so that all
adverse events that occurred in the year after
product was granted exclusivity were reviewed.
Again, this is just to inform you of the ongoing
pediatric activities that are occurring at the FDA,
some of them.
The new advisory committee, I should say
full Pediatric Advisory Committee is meeting for
the first time today. It was chartered this year
and is mandated to include patient and family
organizations, and its mandated responsibilities
include safety, labeling disputes and Subpart D
referrals and general pediatric issues.
The first Subpart D ethics panel met this
past Friday and will report to this committee on
Wednesday. I will tell you a little bit more about
It is important to understand that Subpart
D, which is part of the Common Rule, was
extra protections for children applied to only
federally funded activities until recently. In
2000, the Children's Health Act required FDA
regulated products to be in compliance with
additional protections for children that are
embodied in the Subpart D of the Common Rule.
Subpart D is fundamentally a referral
process. There is much more to it but it is a
process for IRBs when they are unsure they can
approve or under which regulation they should
approve a study involving children. The Pediatric
Ethics Subcommittee reports to the Pediatric
Advisory Committee and this is a public process.
The disclosure of the studies that are
conducted in children is distinct for studies that
are conducted in children under the exclusivity
provisions of BPCA. I mention this because it is
unusual in the FDA if a product is not approved
that those studies would be disclosed. However, we
now have, again under BPCA, a requirement that
within 180 days of submission of a pediatric study
a public summary of the medical and
pharmacology reviews will be posted. There are now
41 pediatric summaries posted at this website.
Basically, you can go to the FDA page and get there
by going to the Center for Drugs or pediatric
The summaries of Effexor, Paxil, Serzone,
Celexa, Zoloft and Remeron are available on the
pediatric summary review site. As you know, and
will hear, Prozac is the only antidepressant that
is approved for use in children, and it is posted
up on FDA's site for approved applications. That
URL is provided for you here and in your handouts.
The new pediatric data has taught us that
our knowledge of pediatric therapeutics is in its
infancy; that we must study children if we are to
understand pediatric-specific adverse events and
reactions or if a product is going to work in
children. The pharmacokinetics in children has
proven to be more variable than anticipated. The
submitted studies that we are receiving are
teaching us that we need to know more about
pediatric endpoints, pediatric trial
how to conduct these trials, and that we will need
to change some of our trials as we move forward.
Ethical issues require reassessment from a
pediatric perspective. Therefore, at this point no
longer shall each child be an experiment of one in
which not much knowledge is gained.
As we move forward, it will require our
careful attention if we are to discover why
children are behaving differently. If children are
to be appropriately treated, we will need to know
more than how to correct those things or describe
adverse events. We are going to need answers to
such fundamental questions as to why children react
differently, what are the metabolic, physiologic
events that are occurring that necessitate
different dosing, or why is there a therapy that
works in adults and does not work in children. Our
public policy must be more knowledge to replace our
ignorance. Thank you, and we look forward to your
DR. MARANGELL: Dr. Murphy, a quick
question, when the FDA requests a study
specify methodology and assessments that you would
like to see included?
DR. MURPHY: When FDA requests a study we
do put in that written request the trial design,
the number of patients, the adverse events--you
know, under exclusivity all of that does go into
the written request.
DR. GOODMAN: Thank you, Dr. Murphy. Now
Overview of Issues
DR. KATZ: Thank you, Dr. Goodman, and
good morning. I would like to welcome you to this
joint meeting of the Psychopharmacologic Drugs
Advisory Committee and the Pediatric Advisory
As you know, we are here to present to you
and to ask for your guidance in interpreting the
results of our analyses of the relationship between
antidepressant drug use and suicidal behavior in
controlled trials in pediatric patients. This
meeting is in follow-up to the meeting of these two
committees held in February of this
year. At that
meeting, as you recall, we presented to you and
obtained your endorsement of our plans to perform
At this point I would like to very briefly
recap how we arrived to this point. As you know,
we first became aware of a possible relationship
between antidepressant drug use and suicidal
behavior in pediatric patients in May of 2003 when,
in response to our request for further
clarification of their data, GlaxoSmithKline, the
manufacturer of Paxil, submitted data to us that
suggested such a link for that drug. As a result
of this submission, the agency issued a public
statement recommending that this drug not be used
in pediatric patients with depression, and
independently we asked all other manufacturers of
antidepressant drugs to resubmit the relevant data
from controlled studies with their drugs in
Based on our review of this data, we
issued a statement informing prescribers that there
was a potential relationship between all
drugs and suicidal behavior, and that these drugs
should be used with caution in these patients.
However, at that time we also noticed that
the data submitted to us from the various companies
was not reported to us in a form that would permit
definitive analyses. Specifically, each company
classified various behaviors as being
suicide-related adverse events in their own
idiosyncratic manner. This led to questions about
whether or not these events were, in fact,
suicide-related and, in addition, prevented
meaningful comparisons between drugs in this class.
For this reason, we decided that an
independent assessment of these possible events by
experts in suicidology would be the most
appropriate way to definitively answer the question
of whether or not any, all or none of these drugs
increased the risk of suicidal behavior in
pediatric patients. Let me just add that by
definitive analyses I mean analyses that make the
best possible use of the available data.
It was at this time that we
issue before you. At that meeting we presented you
with our plans to submit blinded narrative
descriptions of possible suicide-related events to
a group of independent experts, to be coordinated
by Columbia University, whose task it would be to
classify these events as being suicide-related or
not. Although no formal vote was taken, this
committee fully endorsed this effort and agreed
that the data in hand at that time did not permit
definitive analyses to be done.
The committee also recommended, based in
part on the data in hand but also, I believe
importantly, on the basis of testimony from members
of the public who had suffered the tragedy of loved
ones who had committed suicide while taking these
drugs, that the agency should ask sponsors of these
products to warn prescribers that patients being
treated with these drugs, especially at the
beginning of treatment, should be closely watched
for the emergence of signs and symptoms that might
suggest a worsening in their clinical state.
Since that February meeting a
important things have happened. Based on your
advice, the agency drafted, and all of the sponsors
of these drugs have adopted, language in product
labeling warning about the possibility of
significant behavioral changes at the outset of
treatment with these drugs in both pediatric and
adult patients, and the prescribing community and
the public have been informed of these changes.
Critically, this warning made clear that
the possibility of worsening and a possible
increased risk of suicidal behavior at the outset
of treatment could not necessarily be attributed to
the drugs because the data did not permit such a
definitive conclusion. Nonetheless, it was
considered appropriate and prudent to inform
prescribers and patients and their families that
changes in behavior could occur with the onset of
Also, the Columbia group has completed
their task of reclassifying the potential cases of
suicidal behavior and, importantly, we have
completed our reanalyses of these
promised at our February meeting, we are now ready
to present to you the results of these analyses.
At this point I would just like to give
you a brief overview of the agenda for today's and
tomorrow's session. First Dr. Tom Laughren, of the
Neuropharmacology Division, will provide you with a
more detailed account of the regulatory history and
events that have brought us here this morning. He
will be followed by Dr. Diane Wysowski, of the
agency's Office of Drug Safety, who will briefly
present the results of some recently published
epidemiologic studies relevant to this question.
We have provided the committees with copies of
these published materials. Although, of course, we
consider our reanalyses of the controlled data to
be the primary source of data on which your
discussions and recommendations will be based, we
thought it important to present at least briefly
the available relevant epidemiologic data.
Next, Dr. John March, of Duke University,
will present a brief report of the Treatment for
Adolescent Depression Study, or TADS
recently completed trial that evaluated the effects
of fluoxetine in adolescents with depression. As
you know, fluoxetine is the only drug approved in
the United States for the treatment of depression
in pediatric patients and, as you will see, these
data make an important contribution to our overall
assessment of the problem before us.
Dr. Greg Dubitsky, again of the
Neuropharmacology Division, will then present an
overview of the design of the pediatric trials from
which the data for our analyses were derived. This
exploration is important because similarities and
differences in design elements among these trials
can have important implications for whether or not
these data can be examined as a whole, or whether
they must be considered separately.
Then, Dr. Kelly Posner, of Columbia
University and the primary person responsible for
coordinating the blinded reclassification effort,
will present to you the methodology her group used
to produce what we now consider to be the
definitive data on which we have based
Because the reclassification of these
clinical events was the critical activity on which
all subsequent analyses and decisions will have
been based, and because by its nature it involved
subjective assessments of the primary data, we felt
strongly that it was appropriate to ensure that the
methodology used by the Columbia group could
reliably and reproducibly yield similar results
when applied by an independent group. For this
reason, agency scientists performed such an
independent reclassification of a percentage of
these cases, utilizing the Columbia classification
schema, and Dr. Solomon Iyasu, of the agency's
pediatric group, will present the results of this
independent audit of the Columbia process.
At that point, Dr. Tarek Hammad, of the
neuropharmacology group, will then present the most
critical results of his extensive reanalyses of the
data as reclassified by the group of outside
experts. These analyses look at the data for
individual drugs, as well as across all
will provide the data on which the committee
subsequent discussions will be based.
Finally, the last formal agency
presentation will be given by Dr. Andrew Mosholder,
of the Office of Drug Safety. Dr. Mosholder's name
is undoubtedly familiar to you. Dr. Mosholder was
the agency reviewer who had, prior to the February
advisory committee meeting, performed analyses on
the cases as submitted, that is, the
non-reclassified cases, and had concluded that
these drugs did, in fact, increase the risk of
suicide-related behaviors in this population. As
you know, Dr. Mosholder did not present his
conclusions at the February meeting, although the
data on which his analyses were based were
presented and we noted at that time that some in
the agency had already reached a definitive
conclusion on this question.
There has been since that meeting
considerable public discussion and controversy
related to the fact that Dr. Mosholder was not
given the opportunity to present his
that meeting. The reasons for our decision at that
time were straightforward. As I have discussed
today and as we have discussed publicly on numerous
occasions, we had decided that at the time of the
February meeting the data had not been submitted in
a form in which we could reliably agree that the
events described as representing suicide-related
behavior did, in fact, represent such behavior.
We, therefore, felt that conclusions reached on the
basis of analyses that relied on these descriptions
could no, in turn, be considered completely
We felt, and still feel, that presenting
conclusions based on potentially unreliable
analyses could have led to errors in either
direction, that is, resulted in a conclusion that
the drugs were dangerous when they really were not,
or resulted in a conclusion that the drugs were
safe when they were not. A mistake of either kind
could have, in our view, disastrous consequences.
For this reason it was, and remains, our view that
these decisions must be based on the most
analyses possible. Now that the definitive
analyses have been done, however, Dr. Mosholder
will present his own analyses and conclusions, with
particular attention to a comparison between his
results and Dr. Hammad's.
Following lunch we will hear brief
comments from several of the pharmaceutical
companies who have antidepressant drug products on
the market, and the day will end with the open
public hearing. A total of 73 members of the
public have signed up to make statements. As you
have heard and as in the February meeting, we will
again need to limit the statements from the public,
this time to three minutes per individual. We
recognize that this is not much time and we
apologize for the limit but it would be impossible
for all those who wish to make statements to do so
without imposing this limitation. We appreciate
your understanding on this point and, as you have
heard, anyone who wishes may submit written
testimony to the docket.
Tomorrow the committee will
data you will have heard. We, of course, look
forward to this discussion and in particular to
your answers to the questions we have brought to
you and which we have provided in your background
packages. We are, in brief, interested in your
views on our approach to the reclassification
effort and, critically, whether you believe that
the analyses establish that one or more of the
drugs studied increases the risk of suicidality in
pediatric patients. Importantly, if you do
conclude that there is a signal for suicidality,
whether for one or more of these drugs, we need to
know what additional regulatory action, if any, you
believe should be taken.
The results of our reanalyses are complex
and their interpretation is not immediately
obvious. They raise difficult questions, not only
about the fundamental meaning of the results of the
analyses for each drug, but also about the
comparability of the various treatments and,
therefore, whether it is appropriate to consider
the drugs as a class for which any
reached should globally apply or whether the drugs
must be considered individually. Further, the
question of any further regulatory action is also a
thorny one and must take into account the
consideration of the lack of available
effectiveness data for all of the drugs, except
fluoxetine, although the absence of this
effectiveness data is not easily interpreted
Because of the complex nature of the
evidence and because of the extraordinary
importance for the public health of the decisions
that we need to make, we are turning to you, the
experts, for guidance on these matters. We thank
you in advance for your efforts.
DR. GOODMAN: Thank you, Dr. Katz. I
would like to invite Dr. Tom Laughren to come to
Regulatory History and Background
DR. LAUGHREN: Thank you, and I would also
like to welcome everyone to the meeting today. I
am going to begin by briefly giving an
events leading up to today's meeting. I am then
going to talk about the key elements in the
division's exploration and analysis of the
pediatrics suicidality data. I will then spend a
little time talking about our March 22nd public
health advisory and the subsequent labeling changes
that have now been implemented. Then I am going to
spend a little time talking about the effectiveness
data. I did this at the last meeting; I will do
this again because I think it is important to have
these data in mind since they are an important part
of the context of this discussion about pediatric
suicidality. Then I am going to quickly go over
the questions and the issues for which we are going
to be seeking feedback tomorrow. I think it is
important that you have these questions in mind as
you hear the talks this morning.
This slide lists a number of the people at
FDA who have been involved in looking at these
data. As you can see, these people come from
various sections of the agency. It is a long list,
and really the point of this slide is
that we take
this matter very seriously and we have invested a
lot of effort into trying to understand these data.
You heard earlier about the two laws,
FDAMA and BPCA, that give FDA authority to grant
additional market exclusivity for companies which
do pediatric studies. The point of this slide is
that most of the data that we are dealing with this
morning come from these types of studies, in other
words, studies that were done to obtain additional
marketing exclusivity. However, we have also
included in our analysis data from a ninth
antidepressant drug, Wellbutrin, that was not
studied for exclusivity. That was one study in
ADHD. We are also including in our analysis data
from the TADS trial that you will hear about in
more detail later in the morning from John March,
As Dr. Katz pointed out, this issue first
came to our attention based on a review of the
Paxil supplement. In that review, the reviewer
noticed that events suggestive of possible
suicidality were subsumed, along with
behavioral events, under the preferred term
"emotional lability." This led FDA to issue a
request to the sponsor, GSK, to explain this coding
practice. Ultimately, that resulted in a report to
FDA, in May of last year, on pediatric suicidality
with Paxil. As Dr. Katz pointed out, that report
did suggest a signal of increased suicidality in
association with drug use, particularly in one of
three depression trials in that program.
What I am going to do in this slide is
very quickly run through subsequent events that led
us up to the February advisory committee meeting.
So, in June of last year we issued a public
statement cautioning about the use of Paxil in
pediatric patients with depression. In July we
issued requests to sponsors of eight other
antidepressant products to ask them to give us the
same kind of summary data that GSK had provided for
In September of last year we held an
internal regulatory briefing at FDA. The purpose
of this was to brief upper management
signal. The two points that I took away from that
meeting from the standpoint of the division's work
were, number one, there was general agreement that
it would be important to try and classify these
events since many of them were not clearly related
to suicidality and we felt it would be very
important to do a rational classification.
Secondly, there was sentiment that we ought to try
and obtain patient-level data information, beyond
the summary information, in order to try and
explain differences among trials and between
In September and October we began to get
responses to our July requests. Also, in October
we issued requests to sponsors for the
patient-level data sets that I mentioned earlier.
Also in October, we decided to go outside of FDA to
get a classification of these cases accomplished.
Then, again as Dr. Katz mentioned, in October we
issued a second public health advisory, this time
extending the cautionary language to all current
generation antidepressants. Finally, in November
and December, having looked at the responses to the
July request for summary data, it occurred to us
that we may not have obtained all of the relevant
events and so we sent additional requests to have a
broader search for events that we would then try
and get classified.
That brings us up to the February advisory
committee that we held. At that meeting you
advised us to basically continue with our analysis
of the data but, in the meantime, to go ahead and
make some labeling changes. In March of this year
we issued a public health advisory announcing the
changes that we had requested. In the meantime,
the classification of the cases was ongoing by the
Columbia group. Those were completed in June of
this year. Then, in August of this year we
completed our analysis of the pediatric suicidality
In this slide what I am doing is basically
summarizing what I think is the major contribution
of the division to this effort. Again, we went to
a lot of effort to try to ensure
case findings, that we had a complete set of events
to have classified. We then worked with Columbia
University to have these events classified.
As an aside, I would like to mention that
this effort, conducted by Kelly Posner and her
group at Columbia and the very exceptional group of
outside experts that they assembled to do this,
represents a very substantial effort that has not
only helped us to understand these data but I think
will have implications for the field in terms of
developing a standard approach to classifying these
kinds of data, and also will lead to guidance
document that, hopefully, will improve
ascertainment for suicidality, which was a very
significant problem in these trials.
Finally, the third effort that we were
involved in was, again, in obtaining the
patient-level data sets that allowed us to try and
explore for confounding and effect modification, in
essence, to try to explain some of the striking
differences we were seeing in the signal across
trials within programs and across
As mentioned, at the February advisory
committee you advised us to go ahead and strengthen
labeling, in particular for monitoring for
suicidality, while we were completing our analysis.
We did this and we announced the request that we
were making in a March 22nd public health advisory.
The changes in labeling that we requested
have now all been implemented for the ten drugs of
interest. I would add here that our plan is to
extend the standard language to all
antidepressants, not just the current generation
and, in fact, that has already been done for some
of these drugs. We are waiting to do it for the
others until we work out the final standard
language, which will be based on advice we get from
you at this meeting.
What I want to do in this slide is to very
quickly go over the labeling changes that have been
implemented now. This slide focuses on the advice
for clinicians who are using antidepressants for
treating any condition really, whether in adult of
pediatric patients. So, the advice is as follows,
first of all, we are asking clinicians to closely
observe patients who are being treated with
antidepressants for clinical worsening and for the
emergence of suicidality, especially at the
beginning of therapy but also at times of dose
Secondly, we are asking clinicians to
consider changing the therapeutic regimen in
patients whose depression is either persistently
worse or whose emergent suicidality is severe,
abrupt in onset, or was not part of the patient's
Finally, we are also asking clinicians to
observe for the emergence of other symptoms as
well, for example, anxiety, agitation, panic
attacks, insomnia, irritability, hostility,
impulsivity, and so forth. The idea here is that
there is a belief, not really solidly empirically
established but a belief that many of these events
may represent precursors to emerging suicidality.
So, we are also asking clinicians to be alert to
This slide focuses on advice for families
and caregivers that also is included in labeling.
We are asking those folks to also be alert to the
emergence of these same symptoms and to report
those symptoms to healthcare providers if they
Now I want to turn to briefly describing
the efficacy data for the 15 short-term trials that
we looked at in our review of these pediatric
supplements. I am going to be focusing on primary
outcomes in those trials. I also want to spend a
little time talking about the difficulty in
interpreting negative findings in this setting and
again I want to note that although I am not going
to be talking about the TADS efficacy data, you
will be hearing about the TADS efficacy data from
John March a little bit later in the morning.
This is kind of a busy slide but basically
each row in this table represents a different
trial. Again, there was a total of 15 trials.
This is color-coded so you can separate the
different programs. There were seven programs.
Paroxetine had three trials. The rest all had two
trials. The column to look at is the far column
where I have summarized the results on the primary
Basically I have characterized the results
as follows: Where the p value on drug versus
placebo on the primary endpoint was less than 0.05,
I am calling it positive. As you can see, that
applies to the two fluoxetine trials and one of the
citalopram trials. If the p value fell between
0.05 and 0.1 I am characterizing it as a trend.
That applies to one of the sertraline trials and
one of the nefazodone trials. If the p value was
greater than 0.1 on that primary endpoint I am
characterizing it as negative. That applies to all
the remaining trials. So, basically what you have
here is three out of 15 trials meeting FDA's
standard for being positive.
The other point I want to make on this
slide is that this represents FDA's view and I
think it is a reasonable standard, however, it is
not the only standard. To illustrate that, I want
to talk about two published papers, one for study
329, the paroxetine trial, a paper that was
published by Keller in 2001. That paper
characterized that trial as a positive study, the
argument being that although it failed on the
primary endpoint it succeeded on all the secondary
endpoints. So, the authors of that paper
considered it a positive trial and many in the
community also considered that a positive study.
Secondly, there was a paper published on
the two sertraline trials by Wagner et al., in
2003, that was based on a pooling of the two
trials. Individually those trials did not make it
but if you pooled them you got a significant p
value. Again, many in the community view that as
evidence of effectiveness of sertraline in
pediatric depression. This does not meet FDA's
standard but the point is that different folks have
different views of the same data.
Now I want to talk a little bit about the
problem of interpreting negative findings in this
First I want to turn to adult depression
trials for drugs that we believe work. Llooking at
trials that on face should work, about half the
time those trials fail. If that failure rate can be
extrapolated to the pediatric population, the
expectation in two study programs and most of these
programs were two study programs--the expectation
is that three out of four times you would fail to
get two positive studies. So, perhaps it shouldn't
have come as such a surprise that many of these
programs failed. On the other hand, the fact that
the overall success rate, again according to FDA's
standard, is only 20 percent success is clearly a
Other factors to think about in looking at
negative trials in this setting is, first of all,
the history of antidepressant trials in pediatric
depression. If you go back to the tricyclic era,
there were 12 trials comparing tricyclics with
placebo in this population. All of them failed.
There are many interpretations of that. One, of
course, is that these drugs simply don't work in
Another might be that there is even
greater heterogeneity in this population of
patients captured under the diagnostic criteria for
major depression than we see in adults. That would
work against getting positive trials.
Another factor to think about is the
somewhat unusual regulatory context in which these
studies were done. Ordinarily, when companies do
studies they only benefit if they get a positive
trial. In this setting they would win in terms of
getting exclusivity whether the trial succeeded or
failed. I don't know whether or not that was a
factor in the conduct of these trials but it is
another thing to think about.
Finally, at the time that we issued
written requests for these programs we were not
routinely asking for phase 2 dose-finding studies
as we are now. That, again, maybe a factor. It is
possible that the dose was not right in some of
In any case, the bottom line in terms of
efficacy is that I think there are plausible
reasons for failure to find efficacy other than the
obvious one that maybe the drugs don't work. On
the other hand, a very important point I believe is
that even though most of these programs have failed
to meet FDA's standard for approval, this is not
the same thing as saying that we have proof that
the drugs have no benefit. The drugs may have
benefit that has simply not yet been demonstrated.
On the other hand, the failure to demonstrate
benefit clearly is a concern, especially when we
have a risk, as we have now seen, of emerging
suicidality. So, the burden is clearly on those
who believe that these drugs do have benefit to
show that benefit. Tomorrow I am going to talk a
little bit about some possible designs for looking
at longer-term benefits with these drugs.
Now what I would like to do is quickly
move through the questions that we are going to be
asking you to discuss and comment on tomorrow.
Again, we think it is important that you have these
in mind as you hear the presentations this morning.
First of all, we are going to
ask you to
comment on our approach to classifying the possible
cases of suicidality and our subsequent analyses of
the resulting data for the now 24 trials--again,
the additional trial is the TADS trial.
The question then would be do the
suicidality data from these trials support the
conclusion that any or all of these drugs increase
the risk of suicidality in pediatric patients? If
the answer to that question is yes, to which of
these nine drugs does that risk apply? In other
words, is this a class effect of all
antidepressants? Does it apply to certain
subclasses within this broader class or only to
If you believe there is a class risk or a
risk that applies only to certain drugs, how should
this information be reflected in the labeling for
each of these products? What, if any, additional
regulatory actions do you think we need to take?
Finally, again we would like you to
consider what additional research might be needed
to further delineate the risks and the
these drugs in patients with pediatric depression?
Thank you very much.
DR. GOODMAN: Thank you, Tom. I imagine
people have questions but I want to try to catch up
this morning to make sure that we have time for all
the presentations. So, I will ask you to hold your
questions and I would like to invite the next
speaker, Dr. Diane Wysowski, who will be looking at
data from different sources other than clinical
Recent Observational Studies of Antidepressants
and Suicidal Behavior
DR. WYSOWSKI: Good morning. In this
presentation I will be reviewing recent studies of
antidepressants and suicidal behavior and briefly
discuss their methods, results and limitations.
I reviewed two types of studies,
ecological and patient-level controlled,
observational studies. Ecological studies show
increasing antidepressant use and simultaneous
decreasing suicide rates. However, such
correlations do not necessarily imply
Findings of ecological studies can be merely
coincidental. Numerous factors such as changes in
risk factors, social and economic changes, more
available counseling, changes in gun access and
choice of a less lethal means of suicide in
children and adolescents may coincide with
decreases in the suicide rate in children and
Ecological studies don't show which factor
or factors are responsible for an observed trend.
Furthermore, an increased relative risk of suicide
with antidepressants in children and adolescents
may coexist with a decreased suicide rate. To
better examine causality, we turned to
patient-level controlled studies, such as
observational studies, in clinical trials.
For the rest of this presentation I will
be focusing on two patient-level controlled,
observational studies. The first is the Jick study
that was published this July in The Journal of the
American Medical Association. It is a matched case
control design based on patient
diagnoses obtained from the United Kingdom's GPRD,
the General Practice Research Database, for the
period 1993-1999. The GPRD is a database of
medical records from general practitioners of more
than three million patients in the United Kingdom.
For this study subjects were 10 through 69
years of age. Exposures studied were the most
widely used antidepressants in the U.K.,
amitriptyline, fluoxetine, paroxetine and
dothiepin. Dothiepin was chosen as the reference
category. From data on these antidepressants
users, the investigators identified 555 cases of
nonfatal suicidal behavior, defined as ideation or
attempts. They identified 17 cases of suicide.
From the base group of antidepressant users, the
investigators matched the cases with more than 2000
controls who did not develop suicidal behavior.
The researchers then compared the suicidal cases to
the non-suicidal controls for initiation of each
Controlling for age, sex, calendar time
and time from first antidepressant
onset of suicidal behavior, the range of relative
risk for nonfatal suicidal behavior was 0.83 to
1.29 for the antidepressants compared to dothiepin.
None of these risks were statistically significant.
Paroxetine, with a relative risk of 1.29 and a 95
percent confidence interval of 0.97 to 1.7, had
borderline statistical significance.
Similar results were obtained for those
10-19 years old. No statistically significant
association was found between each antidepressant
and completed suicide. No statistically
significant association was found between stopping
an antidepressant and nonfatal suicidal behavior.
The relative risk for nonfatal suicidal
behavior and suicide were highest for patients
first prescribed an antidepressant within 1-9 days,
versus 90 days or more, before the suicidal
behavior of the case in the same time period for
the control. For nonfatal suicidal behavior the
relative risk for antidepressant use within 1-9
days was 4, with a 95 percent confidence interval
of 2.89 to 5.74. For suicide the relative risk for
antidepressant use within 1-9 days was 38, with a
wide confidence interval of 6.2 to 231.
Reviewing the limitations of this study,
the results are only as good as the GPRD data.
There are concerns about possible missing data,
possible incomplete ascertainment and
misclassification of patients, and possible
uncontrolled biases among the antidepressant drugs,
such as selection by severity of depression. There
were no interviews of cases and controls so
medication compliance is not systematically known.
There was no unexposed group, and the
antidepressant risks are only in reference to the
FDA asked Dr. Jick and colleagues to
reanalyze their results with amitriptyline as the
reference category. They kindly responded to our
request, and asked that their interpretation of the
results be presented verbatim to the committee. If
the committee wishes to see these supplemental
analyses I will be glad to present them in the
question and answer period.
Other limitations of the study include the
fact that suicidal ideation is a more subjective,
softer diagnosis than suicide attempts and the
risks were not examined separately. This was a
study of mostly adults and there is limited
information on children and adolescents.
Finally, the investigators excluded
patients with a history of 11 other
neuropsychiatric diagnoses, calling into question
the representativeness of the patients compared
with those in clinical practice.
Another patient-level controlled study
examined the relationship between antidepressants
and the risk of suicide attempt by adolescents with
major depressive disorder diagnoses. The study was
done by investigators at the University of Colorado
School of Pharmacy and Medicine. Robert Valuck was
the principal investigator. It was presented as a
poster at the International Society of
Pharmacoeconomics and Outcomes Research meeting,
this past May.
It is a retrospective cohort
study of paid
medical claims data from the PharmMetrics
Integrated Outcomes Database of 70 managed health
plans for the period 1995 through March, 2003.
Paid claims data include health care provided in
which costs are incurred, such as for
prescriptions, doctor visits, emergency room visits
The investigators identified about 16,000
adolescents aged 12-18 with the first major
depressive disorder diagnosis. They classified
patients into cohorts by antidepressant
prescription, those who received none over the
entire follow-up period, which was the reference
group, and those who received SSRIs, tricyclic
antidepressants or other antidepressants within 30
days of diagnosis. They followed the cohorts for
at least 6 months.
The researchers used a Cox proportional
hazards regression analysis to control for some 14
covariates and to examine the multivariate
relationship between antidepressant use and time to
suicide attempt. The majority of patients, 78
percent, had no antidepressant filled in the 6
months after diagnosis; 15 percent had SSRIs filled
within 30 days of diagnosis. And, 209, 1.3
percent, of the 16,000 patients made at least one
suicide attempt in the follow-up period.
The investigators concluded that
antidepressant treatment with any class of drugs
did not increase the risk of suicide attempt.
Antidepressant use for less than 6 months, compared
with use for 6 months or more, was associated with
a statistically significant 3-fold increased risk
of suicide attempt. Females, those who received
psychotherapy within 90 days of major depressive
diagnosis, patients with substance abuse,
schizophrenia or another mental health disorder,
patients with more chronic diseases and those in
the Midwest and West were independently at greater
risk of suicide attempt.
Dr. Valuck and co-investigators recently
expanded their study to include 24,000 eligible
patients with a new diagnosed major depressive
This expanded study is currently being
reviewed for publication. The researchers added a
propensity matching adjustment to control for
predictors of treatment and to achieve greater
balance among the antidepressant groups. The
proportion of suicide attempters, 1.4 percent, was
about the same proportion as in their smaller
In this expanded study the hazards ratio
for SSRIs compared to no treatment was 1.58, not
statistically significant. The hazards ratio for
tricyclics was not estimable due to small numbers
and it was 1 for the other antidepressant category.
The hazards ratio for multiple antidepressants was
1.43, also not statistically significant. The risk
of suicide attempt declined with longer use of an
antidepressant. Compared with patients having less
than 8 weeks of use, those with equal to or greater
than 6 months of use had a statistically
significant decline in the risk of suicide attempt.
Concerning the limitations, the results
are only as good as these paid claims data. There
are concerns about possible missing data,
incomplete ascertainment and misclassification of
patients, and possible uncontrolled selection
biases by antidepressant group. There were no
interviews of patients so we don't have
systematically collected information on medication
compliance. There are no data on the outcomes of
the attempts. We don't know how many of the
attempters died, and there is no information on
suicides. Also, there is no information on
individual antidepressants. There were differences
in study results between the poster and the
expanded study, although this is probably due to
the larger size of the expanded study.
In conclusion, although most of the
results for the individual antidepressants or
classes of antidepressants were not statistically
significantly associated with suicidal behavior, I
do not believe that the Jick and Valuck studies
completely rule out a possible increased risk of
suicidal behavior with antidepressant use. The
studies reviewed were in agreement in showing that
the risk of suicidal events occurred
significantly closer to diagnosis and onset of
antidepressant treatment. The studies did not
provide data about the characteristics of patients
who did not respond to antidepressants or whose
illness worsened with them. The studies had actual
or potential methodological limitations.
I conclude that more definitive studies,
perhaps large randomized, controlled trials of
sufficient length, are needed concerning the risk
of suicidal behavior and suicide as related to
antidepressant use in children and adolescents.
With so much at stake, children and adolescents,
their parents and physicians and society in general
deserve to know which therapies and which
individuals work best for treatment of depression.
DR. GOODMAN: Thank you very much, Diane.
My preference would be that you present those
additional Jick data tomorrow. I don't think we
have time for it today. Would that be an agreement
by the committee as well? So, I think we are in
accord on that, if you could prepare to
that data tomorrow to us. There is one question,
yes, we will allow that.
DR. PINE: I am wondering if you could
clarify in the Valuck 24,000 patient study, if you
looked at the association in the less than 8-week
treatment versus no treatment group. Did that
confidence interval exclude 1? I mean, I saw that
you gave less than 8 weeks versus prolonged
treatment but I didn't see an odds ratio for less
than 8 weeks versus no treatment.
DR. WYSOWSKI: I don't think that I have
those data. I don't think that they did that
analysis. Their results are still being considered
for publication and we just got an abstract. You
saw the poster in your package. Then, when they
did the expanded analysis they only gave us an
abstract of the results. So, we don't have a lot
of detail on the expanded study.
DR. GOODMAN: Dr. Fost, you had a question
DR. FOST: One of the theories of why
suicide behavior might be increased
prescribing is that the patients are at the worst
then and that is why they are started on
prescriptions. If that were true, one might expect
to see increased suicidal behavior in the week or
two before prescribing. Do either of these studies
allow for that analysis?
DR. WYSOWSKI: I believe the Jick study
did look at that. Actually, no--no, I don't see
any information on that; it is just after.
DR. FOST: And the data set doesn't allow
itself for that reanalysis?
DR. WYSOWSKI: Well, it may. I don't know
whether either investigator has information on
DR. GOODMAN: I think that was an
excellent question. Now I would like to invite Dr.
John March, from Duke University, to present
results from the TADS trial.
Brief Report on TADS Trial
DR. MARCH: Thanks, it is a pleasure to be
here and I would like to begin the presentation by
thanking the committee for inviting the TADS
to present the TADS data, and also thank the FDA
for the comprehensive and thoughtful way that it is
approaching this question of enormous public health
The TADS trial, the Treatment for
Adolescents with Depression Study, is an
NIMH-funded comparative treatment trial, and I am
going to present efficacy and safety data from the
stage-1 outcomes that were published in The Journal
of American Medical Association several weeks ago.
We have a detailed safety paper in preparation. We
have a methods paper which has been published and a
baseline paper which looks at the sample
composition in press, and those of you who are
interested in the TADS trial are referred to these
papers for further information.
As I mentioned, this is a study funded by
the NIMH, coordinated by the Department of
Psychiatry at Duke University and the Duke Clinical
Research Institute, the DCRI. It has had the
benefit of oversight and consultation from numerous
consultants, a scientific advisory
NIMH, DSMB participants included 12 sites from
around the United States. Lilly provided
fluoxetine under an independent educational grant
to Duke University, had no input into the design of
the study, the conduct of the study, the analysis
of the data or the preparation of the manuscript.
My sense is that the major credit for this work, as
for all of the research on which we base
evidence-based practice, goes to the children and
families who are willing to participate in
research. Without their participation, we would
have no evidence at all.
The overall objective of the TADS trial
was to examine the effectiveness of medication and
cognitive behavioral psychotherapy alone and in
combination for the acute and long-term treatment
for adolescents with DSM-IV diagnosis of major
depression. The design of the trial was a
balanced, randomized, controlled study that was
masked by use of independent evaluators; four
groups, placebo, cognitive behavioral
psychotherapy, fluoxetine and their
and the study involved 36 weeks of treatment, of
which I am going to present the stage-1 data for
the first 12 weeks of treatment. We also have a
year of naturalistic follow-up and we have recently
been funded to follow these youngsters out into
young adulthood. That data will not be discussed
Now, it is important, in understanding the
generalizability of the data, to know a little bit
about the sample composition. The inclusion
criteria included outpatients, both boys and girls
age 12-17, with a DSM-IV diagnosis of major
depressive disorder and an IQ greater than or equal
to 80. Youngsters with severe conduct disorder or
substance abuse, other than nicotine, pervasive
developmental disorders, thought disorder, bipolar
disorder, or history of suicidality or homicidality
were excluded from the trial.
Because suicidality is the question at
issue today, I thought it important to say a little
bit more about these exclusion criteria, kids were
excluded if they had a hospitalization
previous 3 months or if they were considered to be
high risk, which meant a suicidal event of some
sort within the past 6 months, the presence of
active intent or plan, or if they had suicidal
ideation in the context of a family which was so
disorganized that we felt that even with the
intensive monitoring in the TADS trial framework
that it would not be reasonable to enter them into
a randomized, controlled research study.
This was a moderate to moderately severely
ill population. We had 439 kids, as you can see,
randomized equally into the 4 groups. The total
sample on the Children's Depression Rating Scale
had a CDRS scale score of 60. That, again, is a
mean score of moderate to moderately severely
depressed, with a range from mild to severe
depression. The T score for that mean score is 75.
That means that these kids were more than 2
standard deviations out from normal with respect to
severity of depression. The sample was multiply
comorbid, as is characteristic of patients in the
clinical samples. This was the first major
depressive episode for about 90 percent of the
sample. Ten percent of the sample had had more
than one depressive episode. The mean duration of
major depression was 42 weeks. Again, over 50
percent of the sample was comorbid for another
mental disorder, both internalizing and
externalizing disorders; 14 percent of the sample
had ADHD and half of those kids were on concurrent
So, unlike the industry-funded trials, the
TADS sample is largely representative of patients
who are treated in clinical practice with major
depressive disorder. As you might expect, given
the severity of illness and the pattern of
comorbidity, these youngsters suffered significant
functional impairment. This is the children's CGAS
rating and you can see that the mean CGAS score was
between 40-50, so significant functional impairment
associated with mental illness in this patient
What did we learn in the trial? This is a
take-home efficacy message. Four groups, again,
began at a CDRS raw score of 60. These are random
regression analyses looking at the adjusted or
predicted means at baseline, week 6 and week 12 of
treatment. Actually, all 4 treatments showed
significant improvement, a characteristic of major
depression. The placebo group and the cognitive
behavioral psychotherapy group were superimposed,
one on top of the other. There is no additional
benefit from receiving cognitive behavioral
psychotherapy, either in the slope term or at the
end point, over receiving placebo. There was
significant benefit from fluoxetine alone, and the
largest effect was associated with the combination
condition. The combination condition beat the CBT
condition and the placebo condition on all 4
efficacy measures. Fluoxetine beat these CBT on
all 4 measures and placebo and placebo on 3 of the
If we look at the impact of treatment
using effect size calculations, the mean of the
control condition minus the mean of the placebo
condition, divided by the pooled standard
deviation, the effect size for the combination was
close to 1. This is a very large effect. For
fluoxetine it was around 0.6, a moderate to large
effect. For CBT there was no difference between
CBT and placebo, effect size calculated relative to
If we look at response rates, defined as a
clinical global importance measure rated by the
independent evaluator of much improved or very much
improved, 71 percent of the combination kids
improved; 61 percent of the fluoxetine-treated
patients improved; 43 percent of the cognitive
behavioral psychotherapy treated patients and 35
percent of the placebo-treated patients improved.
Combination statistically was no different than
fluoxetine. CBT was no different than placebo.
The two drug-containing conditions were superior to
the two non-drug-containing conditions on responder
If we look at the effect size calculated
as a derivative of the odds ratio of improvement
for the active treatments relative to
the responder analyses, the results parallel the
analyses on a scale or outcome variable, the
Children's Depression Rating Scale. The effect
size for combination was 0.8, almost 0.6 for
fluoxetine, and about 0.2 for cognitive behavioral
psychotherapy. So, again, clear superiority for
the drug-containing conditions, with the largest
effect reserved for the combination of medication
management and CBT.
Now, of interest here are the safety
outcomes from the TADS trial. Although we spent an
enormous amount of time and energy on measuring
adverse events, particularly measuring the impact
of the treatments on the potential for harm, I
think it is important to point out that the study
did not have safety as a primary outcome. With 439
subjects randomized to 4 conditions, it is easy to
see that for these outcomes the study is clearly
It is I think important to separate
ideation from behavior. Despite the exclusion, we
had significant suicidal ideation in the
sample. This is looking at the Reynolds Adolescent
Depression Scale, item 14, and 7.5 percent of the
sample exhibited a score of 4 or above which is the
threshold for clinical investigation on the RADS.
CDRS item 6, serious suicidal ideation, the kind of
suicidal ideation that leads you to consider
hospitalization, 2 percent of the sample met CDRS
item 13 criteria for severe suicidal ideation. On
the suicidal ideation questionnaire 2 measures, the
SIQ flag which is to prompt clinical investigation,
or elevated scores on any one of the items on the
SIQ that would also prompt suicidal ideation, 29
percent and 10 percent of the sample respectively
on these measures exhibited clinically significant
suicidality. So, although suicidality was an
exclusion criterion, there was plenty of suicidal
ideation exhibited in the TADS sample at baseline.
Now, as expected, suicidal ideation
occurred across all 4 groups taken in the
aggregate. One sees here, looking at the CDRS item
13 score, in this case greater than 1, or SIQ
score, SIQ flag greater than 31,
suicidal ideation at baseline. It came down at
week 6 and was significantly reduced overall at
week 12. This is the aggregated data across all 4
Random regression analyses looking at
between group differences on the SIQ, although it
looks like these groups might be different at
baseline, in fact there were no statistically
significant differences on the SIQ in all 4
One sees a different result than we found
in the pattern for depression. This is placebo;
this is fluoxetine; this is CBT and this is
combination. The take-home messages here are
three. First, as we saw in the previous slide,
suicidal ideation improves with treatment
irrespective of which treatment one gets. Second,
fluoxetine and placebo are indistinguishable with
respect to suicidal ideation, either with respect
to the slope or at entry, indicating that
fluoxetine, at least on average, is not provoking
suicidal ideation. Finally, the only treatment
which separated from placebo with respect to
reducing suicidal ideation was the combination of
fluoxetine and CBT. So, here the combination
offers a significant advantage over medication
Moving on to behavior, using a
comprehensive adverse event monitoring procedure,
we looked at the incidence of three kinds of
harm-related adverse events. Harm-related events,
the broadest category, were defined as harm to
self. This could involve no suicidal ideation,
ideation or attempt. Or, harm to others which
required aggressive ideation or actual behavior
involving harming another person or physical
property. These events are subsumed one within the
next. So, a suicide-related event, which is a
subset of harm-related events, involves harm to
self, either ideation or attempt. Then we had
suicide attempts themselves. What differentiates
harm-related events from suicide-related events is
primarily one subject with aggression and 7
subjects, I believe, who exhibited
without ideation, primarily cutting.
These are the actual rates of harm-related
and suicide-related events divided by treatment
group. One sees that there is a larger number of
harm-related events and suicide-related events in
fluoxetine-treated kids relative to placebo-treated
kids. The combination group is intermediate
between harm-related and suicide-related events,
intermediate between fluoxetine and placebo. The
cognitive behavioral psychotherapy group was
roughly comparable to the placebo group.
If you look at children who received drug,
that is, combining the fluoxetine and the
combination groups and comparing them to the
placebo group, 10 percent, 22 of those
fluoxetine-treated kids exhibited a harm-related
event; 7 percent exhibited a suicide-related event;
and the rates of these events overall were quite
low, 7.5 percent of 439 kids, or 33 kids had a
harm-related event, 24 of 439 kids, or 5.5 percent
exhibited a suicide-related event.
I think it is very important as
through this discussion to understand that the base
rates of these events are extremely low relative to
the rates that we see for benefit.
If we look at the odds ratios calculated
from the actual rate data, the relative risk is 1.5
for combination, 2 for fluoxetine, less than 1 for
CBT. Those is calculated relative to placebo. For
the collapsed category of fluoxetine and
combination the relative risk is slightly greater
than 2. This is the only statistically significant
relative risk in which the confidence interval
crosses 1. That is largely because these events
are so rare so the power is quite low to identify
these events. In fact, the power for detecting a
20 percent difference is about 10 percent.
For suicide-related events there are no
statistically significant differences although, as
you can see from the graph, the odds ratios pretty
closely track the odds ratios for harm-related
The take-home message from this
presentation actually is in this
table--no, it is
in the next table. This is the table that looks at
the suicide attempts in the trial. We had 7 of
them out of 439 kids, or slightly less than 2
percent of the sample, Two fluoxetine-treated
kids, 4 combination-treated kids, 1 CBT and no
placebo-treated patients made a suicide attempt.
There probably is an imbalance in randomization
which may in part be responsible for this. There
were more kids with an elevated SIQ flag randomized
to the drug-containing than the non-drug-containing
conditions so it is not clear what to make of this
Here I think is where the take-home
message lies relative to safety. This is looking
at the benefit/risk ratio using analyses for the
number needed to treat and number needed to harm.
What we see here is fluoxetine compared to placebo,
in the first column; combination compared to
placebo; and the collapsed category, SSRI versus no
SSRI. The absolute benefit increase is calculated
as the control, the experimental event rate minus
the control event rate. So, it is the absolute
benefit increase for receiving the treatment. The
absolute risk increase is calculated, again, as the
experimental event rate minus the control event
rate, that is, the risk increase attributable to
the treatment over the placebo condition in
The NNT and the NNH are the reciprocal of
the absolute benefit increase and the absolute risk
increase respectively, 1 over the absolute benefit
increase or 1 over the absolute risk increase.
These are defined as the number of patients for
benefit that would need to be treated to find one
patient who had benefit over the benefit occurring
from the control condition or, in the case of the
NNH, the number of patients who would need to be
treated to find one patient who would be harmed
over treatment with the control condition. So, it
is a nice metric that combines both the absolute
rate and the magnitude of the effect.
One sees clearly here that 27 percent of
patients benefit from treatment with fluoxetine,
with an NNT of 4 which is a large effect;
percent of patients benefit from treatment with
combination, with an NNT of 3, again a very large
effect; for SSRI versus no SSRI, combining the
categories, 31 percent of patients benefit, again
with an NNT of 3.
The absolute risk increase for a suicidal
event with respect to fluoxetine is 4.7 percent as
compared to placebo; 2 percent for combination over
placebo; and 3 percent for the combined category.
NNH is number of patients that you would need to
treat with the active treatment to find one patient
who would be harmed over the control condition of
21, 50 for the combination condition and 4 for the
From a clinical point of view, these
patients would be easy to pick out in a crowd,
easily identifiable who is getting better and who
is not getting better, active treatment versus
control. These effects are so small and so
uncommon that one could not possibly pick out
patients who would be harmed by the medication
versus patients who would commit these
suicide-related event behaviors with placebo
treatment. If you calculate the NNT to NNH ratio
looking at benefit to risk, one sees clearly here
that the benefit tilts in favor of the treatment,
and particularly the combination treatment.
So, we conclude that the combination of
fluoxetine and CBT is the most effective treatment
for adolescents with major depression. Fluoxetine
alone is effective but not as effective as the
combination of the two treatments. CBT alone is
less effective than fluoxetine and not
significantly more effective than placebo. We also
conclude that placebo is acceptable in randomized,
controlled trials of adolescent major depression
and, in fact, is essential for looking at the
adverse event outcomes, at least in this study.
Suicidality decreases substantially with treatment.
The improvement in suicidality is greatest with the
combination and least for fluoxetine alone.
Fluoxetine does not increase suicidal ideation.
Suicide-related adverse events which are uncommon
may occur more often in
patients. CBT may protect against suicide-related
adverse events in fluoxetine-treated patients.
Taking both risk and benefit into account, the
combination of fluoxetine and CBT appears superior
as a short-term treatment for major depression in
Now, the most practical clinical trialist,
the kind of trial models that are used in other
areas of medicine--cardiology, oncology, infectious
disease for example, would much prefer a large
simple or practical clinical trial in 2000 subjects
to a meta-analysis of 10 under-powered subject
trials. So, it is our sense from looking at the
FDA data and also the TADS data that we have a
significant signal for drug treatment relative to
suicidality but the evidence is not conclusive. In
fact, a definitive study has not been done and we
would, as a field and as consumers of this
information, much benefit from a
placebo-controlled, practical clinical trial
comparing fluoxetine to another SSRI, perhaps
sertraline or citalopram. This trial could be run
easily on the child and adolescent psychiatry
trials network, which is a clinical trials network
that we are now putting in place to run these kinds
of trials in the pediatric population. Thank you.
DR. GOODMAN: Thank you, John. My first
question is will you be here tomorrow?
DR. MARCH: No.
DR. GOODMAN: Oh, you won't? That may
affect my subsequent questions because I am sure,
besides myself, there will be a number of questions
for you. I don't know if we have time to take them
all right now. I wonder if there is any other
option, Anuja. What time do you leave today, Dr.
DR. MARCH: Noon.
Committee Discussion on TADS Trial
DR. GOODMAN: I am going to ask a
question. My understanding in looking at your
results is that on the categorical measures of
response fluoxetine is superior to placebo.
However, if you look at a comparison of the mean
scores on the CDRS fluoxetine is not
placebo. Is that correct?
DR. MARCH: The slope term on the random
regression analysis for the CDRS, the p value was
0.08 for fluoxetine versus placebo. Fluoxetine was
statistically significantly different than CBT but
DR. GOODMAN: So, from a standpoint of
FDA, if this trial had been submitted to the FDA
and you didn't have the CBT group, it seems to me
that it might be classified as a negative study.
DR. MARCH: It would have been classified
as a negative study using the CDRS as the primary
endpoint, the slope term.
DR. GOODMAN: Do you have any comments
about the methodology or outcome measures we are
using and whether we are using the most appropriate
ones in your opinion?
DR. MARCH: Well, I think it is actually a
very important question. If you look at the
fluoxetine outcomes on the predicted endpoint, the
week 12 endpoint on the CDRS predicted by the CDRS
slope, on the clinical global improvement
dichotomized and on the Reynolds Adolescent
Depression Scale fluoxetine was statistically
better than placebo. It was simply a near miss on
the slope term, which probably relates to the way
the random regression analyses handle standard
errors. So, my sense of the story that the data is
telling us is that fluoxetine--and also if you look
at the effect size calculations--the story the data
is telling is that fluoxetine is an effective
treatment and it would be a mistake to consider
this a negative trial. On the other hand, the
technical definition used by the FDA would require
that the study be considered negative.
DR. GOODMAN: Dr. Perrin?
DR. PERRIN: On the other side of the
equation, you made comments that the sample was
somewhat different from some of the trials that
have been sponsored by industry. Could you be a
little bit more specific about what the differences
are, and how they might have affected the results,
and what are the implications for meta-analyses of
DR. MARCH: I think it is a very important
question, and we have a paper that is in press in
The Journal of Child Medicine Psychiatry, the
"orange journal," which describes the TADS sample
in some detail and compares the TADS sample to
epidemiologic samples and to treatment samples,
both on the pharmacotherapy side, primarily the
industry data sets, and also the cognitive
behavioral psychotherapy trials, of which there are
13 published at this point. In general, our sample
is not substantially different from either the
epidemiologic or the treatment seeking samples,
with the caveat that we are slightly sicker and
slightly more comorbid, particularly relative to
the CBT samples.
So, given that the range of depression
goes from mild to severe and half the sample is
comorbid, there are plenty of patients in the data
set who resemble the mildly ill patient all the way
up to the severely ill, multiply comorbid patients.
So, I think the result of the TADS trial is
generalizable to the total sample.
With respect to meta-analyses, it would be
better to have a very large sample including all
these variations, but by combining the data sets
one gets a better picture, I believe, of the total
variation in the patient population than simply
using the industry data sets which tend to exclude
the more complicated and clinically ill patients.
DR. GOODMAN: Dr. Newman, did you have a
DR. NEWMAN: Normally when you use a
continuous outcome you have greater power than when
you dichotomize. I assume that is why you
specified that as the endpoint at the beginning of
the trial. A reason why you might not is that the
medication helps maybe a majority but actually
harms a minority and then you could actually see
that if you dichotomize the percent to improve is
statistically significantly greater in the
fluoxetine group but the mean may not be
significantly greater because there are some people
who are harmed and they drag the mean down, whereas
they have no effect on the dichotomized
Did you look to see whether there was evidence that
the variation in the standard deviation in the
effect size differed between the two groups and
might have been greater with fluoxetine?
DR. MARCH: That was actually the point
that I made, that the standard errors are larger in
the fluoxetine-treated group than in the combo or
DR. NEWMAN: Actually, not just the
standard errors but the standard deviation, meaning
that, in fact, there are some people who are harmed
by it and that diminishes the apparent benefit when
DR. MARCH: It would be impossible to look
at the standard errors or the standard deviation
and make a judgment about harm because there is a
fair amount of variability data point to data point
which is intrinsic to the disorder. Some patients
get better; some patients deteriorate.
We do have in the safety paper a whole set
of analyses looking at shifts, which I am not
confident enough in to have wanted to
today. We will try to examine what percentage of
patients are getting worse with respect to ideation
and behavior, and how that relates to treatment
classification and also how it relates to other
adverse events like mania activation, anxiety
disinhibition and so on. I think the secondary
paper is going to shed a fair bit of light on these
DR. GOODMAN: Dr. Pfeffer?
DR. PFEFFER: I have two questions. One
is were there any differential dropout rates in the
samples? This also relates to the question of
compliance in the different treatments. My last
question is these were intent-to-treat analyses?
DR MARCH: All very good questions. The
analyses are all intent-to-treat. Although I
didn't present the data, if we look at observed
cases analyses, those who were still on their
assigned arm at any given assessment point or
completer analysis, the results are exactly the
same. About 10 percent of the kids in each
treatment overall dropped before the
point. Another 10 percent were what we call
prematurely terminated. That is, for ethical
reasons. They received an out of protocol
treatment at some point during the first 12 weeks
of the study. There were no statistically
significant differences across treatment groups in
either the rate of dropping out or the rate of
receiving an ancillary treatment, that is, a
You will see this afternoon when FDA
presents its analysis of the TADS data that the
odds ratios for being harmed by receiving
fluoxetine are greater than we presented, or I
presented this morning on behalf of the TADS team.
That is because the FDA data set excluded those
kids who were prematurely terminated and received
another treatment. That actually represented two
kids in the placebo group and that, in turn,
inflated the odds ratios in the FDA results versus
the TADS results. So, there is some method
variance in there which accounts for the
differences between the two findings.
DR. GOODMAN: Dr. Gorman?
DR. GORMAN: Does analysis of your data
set allow interpretation of the time of onset of
treatment to behaviors that we are studying today?
DR. MARCH: That is a very good question
and, in fact, one that we will address in the
secondary paper. I can tell you that the majority
of the events occurred within the first 6 weeks but
not within the first 2 weeks. But I don't have
that data presented in slide form so I can show it
to you, but it will be in the safety paper that we
are currently preparing.
DR. GOODMAN: So you don't know what the
differential rates are between the groups at this
point, particularly between fluoxetine and placebo?
DR. MARCH: In terms of time?
DR. GOODMAN: Yes, in terms of the early
DR. MARCH: My general impression, looking
at the data, is that the fluoxetine events occurred
early and the placebo events occurred later, which
is kind of what you would expect given
what we know
about the compounds. But I wouldn't want to cite
chapter and verse or have you base your decisions
based on that because we haven't completed the
final analyses of the data.
DR. GOODMAN: Dr. Rudorfer, you will be
the penultimate questioner.
DR. RUDORFER: If I can go back to the
characteristics of the patients for a moment, we
will be looking at a number of studies that were
submitted to the FDA by various sponsors, and it
seems to me that the TADS inclusion criteria go
beyond DSM-IV in terms of length of illness and
degree of dysfunction in various spheres of life.
Could you comment on that?
DR. MARCH: Sure, it is a very good
question. That is, the exclusion criteria included
requirements that were designed to ensure a stable
baseline. So, we required at least six weeks of
mood disorder symptoms that crossed two or three
contexts--home, peers in school--which, of course,
are not required in the DSM-IV criteria. This was
done in part to minimize the chance of a
response and to ensure that we had a sick patient
population that would be both ethical to randomize
and would offer some opportunity for the
combination treatment to separate from the two
I think we actually designed a very good
experiment, and looking at a 35 percent placebo
response rate did exactly what we had intended to
do. But in that sense, this population is sicker
perhaps than what is seen in the industry data sets
and certainly sicker than what is seen in the CBT
data sets on average.
DR. GOODMAN: I will permit two final
questions and that is it, one from Dr. Pine and
then Dr. Fant.
DR. PINE: I want to return to the first
question from Dr. Goodman as far as how this study
would be evaluated from an FDA perspective. My
sense from reading the paper is that there were two
primary outcome measures and three analyses, and
two of the three were positive so that the CGI
analysis done categorically was positive,
analysis done categorically was positive, and it
was only the third analysis, the CDRS continuous
measure, that was not positive.
So, my take on that from an FDA standpoint
of, you know, do the primary outcome measures make
it or not is that it would be closer to positive
than negative. Do I have that right?
DR. MARCH: You have it partially right.
There is a CDRS slope analysis, and whether that is
positive or negative depends on whether you treat
the intercept term as random. I mean, there is a
fair amount of method variance on a subtle level
that can tilt these things one way or the other
when it is a near miss. There is a CDRS endpoint
analysis based on the predicted or marginal mean.
There is a categorical analysis, logistic
regression, and there is a self-report scale which
was included, the Reynolds Adolescent Depression
Scale, which was included because the CBT
literature relies heavily on patient self-report.
Three of those four measures, all but the CDRS
slope analysis, were positive for the
fluoxetine-placebo comparison. The CDRS slope
analysis, again, was a p value of 0.08, a near
So, I think the take-home message is
actually in the effect sizes, not in whether you
are looking at p values or not. Clearly,
combination and fluoxetine have larger effect
sizes, meaningful effect sizes relative to placebo
as contrasted to CBT.
DR. FANT: For the sake of the study I
know it was necessary to exclude certain patients
to optimize the conditions for the study, but I
think in real-world practice a lot of the patients
who were excluded would be patients who would be
prescribed medication under various conditions. Is
there any reason, from your standpoint or
perspective, to think that that population of
patients may be at a different risk for fulfilling
suicidal attempts, ideation, and carrying it
through to the ultimate result, or may be affected
differently by the medication than patients that
are not excluded from the study?
DR. MARCH: That is a very good question.
That is, is there some issue to believe that there
would be a differential treatment response in
patients who would be excluded, particularly
excluded for harm to self or others, as compared to
the TADS sample of patients? I don't know of any a
priori reason to believe that there would be a
differential treatment effect relative to the TADS
sample. I do think it is quite clear from the
treatment and epidemiologic literature that they
would be at higher risk for adverse harm-related
outcomes but whether they would be more at risk
than, say, the TADS sample patient I don't think we
know. My guess as a clinician is probably not.
We are actually doing an NIMH-funded study
called the Treatment of Adolescent Suicide
Attempter Study, in which we are comparing a
medication algorithm to cognitive behavioral
psychotherapy to the combination--no untreated
control condition obviously in this sample--to try
to understand something about treatment for this
particular patient population precisely
they have been excluded from these other trials,
and we need additional data on their care. So,
that trial is now under way and should be completed
in the next couple of years.
DR. GOODMAN: Thank you very much, John.
The final question that I will take the chairman's
privilege to ask is do you have data that you
haven't presented yet on long-term outcome based on
DR. MARCH: The final subjects in the
trial are out in a naturalistic follow-up window so
that 36-week data is in the can, but that data set
has not been cleaned and locked yet. We expect it
will be cleaned and locked and ready for analysis
in the spring, and we hope to have that data in
press by this time next year.
DR. GOODMAN: Thank you very much, John.
DR. MARCH: Thank you.
DR. GOODMAN: I would like to ask our next
speaker to come forward, Dr. Greg Dubitsky.
Characteristics of Pediatric Antidepressant Trials
DR. DUBITSKY: Good morning.
heard about the TADS trial from Dr. March. I would
like to now go on and briefly summarize the other
studies that were included in the FDA's primary
analysis of suicidality.
I do want to emphasize that my review and
this presentation are really descriptive only. I
am not going to touch on efficacy outcomes or
safety outcomes. The discussion of the risk of
suicidality in these trials with be presented later
this morning by Dr. Hammad.
The study pool, again excluding the TADS
study, consisted of 23 placebo-controlled studies
which were conducted between 1984 and 2001. Each
study was done with one of nine different
antidepressant drugs and studied patients with one
of five different diagnostic indications; major
depression, obsessive compulsive disorder,
generalized anxiety disorder, social anxiety
disorder or attention deficit disorder.
These studies all had some features in
common. They were all randomized, double-blind,
placebo-controlled. They utilized a parallel group
design and a flexible dosing regimen.
I did prepare a handout to go with this
talk which should be in your packets, at least for
the advisory committee members. It consists of two
tables which summarize some of the design
characteristics of these 23 studies. Table 1 has
some basic study information, to include the
diagnostic indication, the age range that was
studied, number of patients by treatment group, the
duration of double-blind treatment, and the dose
range that was used in the particular study.
I would like to point out though that I
don't intend for everybody to read this and
memorize it; this is really for reference for later
this morning when you hear about the analysis of
The second table in the handout includes
some information on screening and exclusionary
criteria. Some of the studies used very extensive
diagnostic screening. I have indicated those in
the table. I have also indicated information on
whether there was a placebo lead-in, and
whether certain exclusionary criteria were employed
in the various studies to include whether people
were excluded who had a history of treatment
resistance, current suicide risk, history of a
suicide attempt, bipolar disorder, or family
history of bipolar disorder.
My review of these studies did include a
number of other variables. I have listed in these
two tables the most relevant ones but in my review
that is on the Internet I do describe some other
characteristics which you might be interested in,
such as the location and number of sites, whether
stratified randomization by age group was utilized
and other exclusionary criteria such as homicidal
risk or the presence of psychotic symptoms.
There were a few notable differences
between these studies that I would like to point
out. One study with Prozac, HCCJ, was a very small
study. It was the smallest of the 23 studies, with
only about 40 patients and it was terminated early.
Only one of the 23 studies included an
active control arm. That was study 329 with Paxil
in major depressive disorder. That included an
imipramine control arm. The others only had a
Two of the studies did include inpatients
as well as outpatients, the Celexa study, 94404,
and Wellbutrin, 75.
Last, I did want to point out that three
of the studies did use a rather extensive
diagnostic screening of the patients, much more so
than the other studies, Prozac studies X065 and
HCJE, and Paxil study 329. Those three studies
were done in major depressive disorder.
One other difference involves the
treatment options after patients completed the
acute phase of double-blind treatment. This was
quite variable across the trials. In eight studies
there was a taper of acute treatment before
discontinuation. Seven other trials just abruptly
discontinued treatment, and there was no provision
for continued treatment. Five trials did allow for
continuation of open-label treatment, and in three
trials patients could continue
However, this was also very variable
within trials. For instance, in Paxil 329
responders could continue double-blind treatment
but non-responders were tapered off treatment.
This variability in the follow-up treatment
following the acute phase made it very difficult to
do any analysis of suicidality-related events post
I would like to point out that none of
these studies was specifically designed to assess
suicidality. Suicide attempts and ideation were
detected only through routine safety monitoring,
that is, through treatment emergent adverse events
and through suicide-related items on various
depression scales, such as the HAM-D and the CDRS.
One problem with this is that often descriptions of
possibly suicide-related events were rather vague
or incomplete and often made it difficult to reach
I have no specific conclusions since this
is really a descriptive review and
overview of the
studies. I think one of the important questions
that arises from this information though is whether
any of these differences in design characteristics
could contribute to any observed differences in
suicidality risk that we observed across these
studies. That is a question that will be addressed
later this morning by my colleague, Dr. Hammad.
So, that is all I have.
DR. GOODMAN: Thank you for being concise
and providing us with an outstanding handout for
our reference. We have one question. Dr.
DR. RUDORFER: Thank you. Could you
clarify, of the 23 trials how many were submitted
in response to the pediatric exclusivity rule?
DR. DUBITSKY: I don't have the exact
number. I believe most of them were but some of
them were submitted well before pediatric
exclusivity took place or came into effect. I
don't have the exact number off the top of my head.
DR. GOODMAN: Dr. O'Fallon?
DR. O'FALLON: Asking the question in a
somewhat different way, the data that you have for
this reanalysis, does any of that data come from
outside, beyond the data that was submitted to the
FDA? That is, were you able to go in and obtain
data from studies that were never submitted to the
FDA for approval or whatever?
DR. DUBITSKY: Well, to my knowledge,
there was one study that had not been submitted as
part of an efficacy supplement or an approval
package. The other ones, I believe, were. Dr.
Hammad actually requested data sets for all these
studies. Correct me if I am wrong, but I believe
we had relatively complete data sets to allow
reasonable analysis for all these studies.
DR. O'FALLON: But I am asking whether
there are, as some are claiming, studies that were
done but were never submitted to the FDA. Are
there any of those data here, if they exist?
DR. DUBITSKY: There are some studies that
are not included in this analysis, but those are
mainly open-label continuation studies of the acute
Also, there were a number of pediatric
pharmacokinetic studies but I think for obvious
reasons we didn't include those in the analysis.
But, to my knowledge, I think we have everything.
DR. GOODMAN: Dr. Laughren, do you also
want to respond to the question?
DR. LAUGHREN: I think I can respond to
that. The vast majority of these programs were
submitted under pediatric exclusivity so the
companies were required to submit every scrap of
data they had as part of those supplements. The
only trial here that was not submitted as part of
an application, in terms of a company trial, was
the ADHD study for Wellbutrin. The other study
that we have included safety data for is the TADS
trial and, of course, that was also independent.
But those are the only two trials of the 24 that we
looked at that were not submitted as part of an
DR. GOODMAN: Dr. Marangell?
DR. MARANGELL: How many of the studies
excluded family history of bipolar disorder?
DR. DUBITSKY: Let's see, actually I think
I have that in table 2. I don't know the number
off the top of my head. It looks like about ten of
the studies excluded a family history of bipolar
DR. MARANGELL: Thank you.
DR. GOODMAN: Dr. Perrin?
DR. PERRIN: You are saying basically that
the extensive diagnostic screening occurred only in
three studies, I believe. Is that right?
DR. DUBITSKY: I am sorry?
DR. PERRIN: The extensive diagnostic
screening occurred only I think in three
studies--one of the points that you made. Does
that give us some information about the potential
diagnostic heterogeneity and also raise questions
about whether entrance into these studies of
children, ages 7-17, might not have had MDD in the
MDD studies? My last related question is, since I
am not a psychiatrist at all, what do we know about
the ability to distinguish bipolar disorder from
MDD in the 7-17 year-olds?
DR. DUBITSKY: Well, it is true that in
looking across all 23 studies, those three studies
did stand out as far as using more extensive
diagnostic criteria. I believe that it certainly
is possible that we might have more confidence that
those patients did actually have the diagnosis
under consideration. Whether that is actually true
or not, I don't know and I don't know any good way
of figuring that out.
I am not a child psychiatrist so I can't
answer your last question about the ability to
diagnose. I understand it is very tricky though.
DR. GOODMAN: Thank you again. I would
like to ask Dr. Kelly Posner to come up to the
podium to present. Dr. Posner is from Columbia
University and she will be talking to us about the
reclassification of the clinical trials data
according to suicidality.
Classification of Suicidality Events
DR. POSNER: I would like to start by
introducing my expert work group from Columbia that
included myself, Dr. Maria Oquendo, Dr. Barbara
Stanley and Dr. Madelyn Gould. Dr. Stanley and Dr.
Gould are here with me today. Our statistical
consultant was Mark Davies.
Why was reclassification needed? The
problem is that the field is challenged by a lack
of well-defined terminology and common language to
refer to suicidal behavior, and this was reflected
in the lack of standardized language used in the 25
trials in question. That is why there was
difficulty in interpreting the meaning of all of
these reported adverse events that occurred in
these trials. So, AEs that should have been called
suicidal may have been missed and there may have
been AEs that were inappropriately classified as
Here are some illustrative examples of the
difficulties in adverse event labeling in the
field. I want to make sure to note that these
labels have nothing to do with the labels the
sponsor gave these events, but just original
investigators at the site. Again, they are extreme
examples just to reflect the problem.
You see the first one, it says
attempted to hang himself with a rope after a
dispute with his father. Investigator did not
consider this event to be a suicide attempt but
called it a personality disorder in this 10
The second one is one we have all heard a
lot about. The patient is reported to have engaged
in an episode of auto-mutilation where she slapped
herself in the face, called a suicide event. Then,
the patient took 11 tablets impulsively then went
to school--called a medication error.
So, how do we address this problem? Well,
a common set of guidelines needed to be applied and
we needed to look at the data consistently across
trials using research-supported definitions and
concepts that had reliability and validity. We
also needed to broaden the range of adverse events
that we were looking at. This was for two reasons.
The first one is to avoid bias in readings. We
wouldn't have wanted the expert raters only to have
had what the sponsors had identified as possibly
Also, to identify suicidal events that
may have been missed.
So, what was included in this broadened
range of events? Of course, the events originally
identified by the sponsors as possibly suicide
related, all accidental injuries which included
accidental overdoses, and serious adverse events
which includes life-threatening events and all
Why did we need experts in suicide? Well,
you all heard about the limited information
provided in the narratives, particularly frequent
lack of stated suicidal intent. So, only experts
in suicide would have allowed for inference based
on details of behaviors and related clinical
This is the list of our very distinguished
international panel of experts. I will just read
their names very quickly, Drs. Bautrais, Brent,
Brown, Van Herringen, King, Mazark, O'Carroll,
Rudd, Spirido and Miller.
So, what was the Columbia classification?
I want to move to this slide because it
through the definitions which I will just go
through briefly. Suicide attempt, of course, which
is defined as a self-injurious behavior associated
with some intent to die. Intent can be stated or
inferred by the rater. It is important to know
that no injury is needed.
Then there was preparatory actions towards
imminent suicidal behavior. So, the person takes
steps to injure himself but is stopped by self or
other, anything beyond the threshold of a
verbalization but not quite making it to a suicide
Then we had self-injury behavior, intent
unknown. These are cases where we know there was
some self-injury but we don't know what the intent
was. So, the associated intent to die is unclear
and cannot be inferred.
Self-injurious behavior with no suicidal
intent is the next category. That is where, again,
we know there was deliberate self-harm but there
was no intent to die so behavior is intended to
affect other things. This is what we think of
Suicidal ideation was the next relevant
category, which can be passive or active thoughts,
passive thoughts of wanting to be dead or active
thoughts about killing oneself.
Then we had all the other categories.
That is essentially one rating, anything other than
deliberate self-harm or something suicidal. That
could include accidents, psychiatric events or
Finally, we had not enough information,
which meant that there was insufficient information
for a rater to be able to say whether or not there
was some deliberate self-harm or something
The scheme is laid out conceptually here
for you. I think it helps make a little more sense
of it. The blue boxes refer to what you will hear
later as the FDA's primary outcome. These are
ratings that are considered definitively suicidal,
suicidal behavior and suicidal ideation. You see
codes 1, 2 and 6. Suicide attempt, preparatory
actions and ideation. The next are non-suicidal
events, all the other events and the self-injurious
behavior without suicidal intent, and then
indeterminants. The green boxes are what will be
referred to as the sensitivity analysis, things
that could have been suicidal but there is no way
So, what was done? The classification
methodology involved, of course, choosing the
expert panel who had expertise in adolescent
suicide and suicide assessment, based on reputation
and publications. They had no involvement in
industry youth depression trials in question, and
no expert rater was employed by Columbia
We had a training teleconference to review
classification parameters, then training
reliability exercise to ensure appropriate
application of classifications. All case
narratives were blinded to any potentially biasing
information, and I will review that in a minute.
There was random distribution of 427
events to 9
expert raters. Each case was independently rated
by 3 raters. Each rater received approximately 125
events to rate, and any group of 3 raters shared
only 5 cases. All ratings were reviewed for
quality assurance and identification of
non-agreement cases. Consensus teleconferences
were held for any disagreement cases, and there was
double data entry for quality assurance.
Now, what was the consensus process I
referred to? If ratings did not have unanimous
agreement, then a consensus discussion was held.
Each case was discussed by the three raters
involved only. Discussion of each case was led by
an expert other than those originally assigned the
case. The goal of the discussion was to reach 100
percent agreement. If 100 percent agreement could
not be reached, the case then became indeterminate.
Sometimes the original majority opinion did not
always end up as the final consensed
classification. In other words, if there was a
minority rating, sometimes that ended up being the
Now, what was rated? Blinding of event
narratives to avoid bias included--we received the
narratives from the FDA blind to all potential drug
identifying information. This included drug name,
company sponsor name, patient identification
numbers, whether they were on an active or placebo
arm, and any and all medication names and types
because there could be associated treatments that
might bias somebody or tip them off as to what drug
was being talked about and, of course, primary
diagnosis. We also did some additional blinding of
potentially biasing information which included the
original label of the event given by the
investigator or sponsor and serious or non-serious
Rating guidelines--how was the
classification scheme applied? We wanted the
experts to apply concepts using their clinical
expertise and judgment; to use their experience to
integrate clinical information and infer when
appropriate. We wanted them to have a reasonable
certainty in order to commit to a rating,
rating was based on what was probable, not what was
The guidelines for intent inference
involved inferring if something was clinically
impressive, and I am going to give you an example
of that in a moment, or using two smaller pieces of
clinical information. The clinical information
that could inform inference of intent included
clinical circumstances. That could be method used,
number of pills; past history of suicide attempt;
past history of self-injurious behavior or
self-mutilation; and family history of suicide or
Now, here is a case example of inferred
intent, what we call clinically impressive
circumstances. This is the first time you are
actually seeing one of these real narratives. In
this case clinical impressiveness actually
overruled stated intent, so you see the subject
attempted suicide by immolation. Her siblings
doused the flames immediately. She was left with
minor burns on her abdomen and on her
shoulder. The subject admitted she was angry with
her parents for going away and leaving her alone at
home because she was fearful. The subject admitted
that she had acted impulsively and had not intended
to kill herself.
Here are more examples. This is another
example actually of clinically impressive
circumstances which was ultimately called a suicide
attempt. It is also important to know that we had
no idea what the sponsor ratings were but both
these cases were consistent with what the sponsor
had said as well.
This case involved a 16 year-old who
claimed to have ingested 100 tablets of study med
after a fight with her mother. The patient
informed her mother. The mother brought her to the
ER. The patient reported feeling shaky. Emergency
room physician said she was slightly tachycardic
with a pulse of 100. The tox. screen was negative
but the patient did have some illness and she
stayed in the ER until she was asymptomatic, and
then was later admitted to the psych. unit.
Another example of a suicide attempt, a
patient age 17 took an overdose of 20 tablets. In
the father's opinion the overdose was 5 tablets.
The patient didn't have any symptoms of an
overdose, not even nausea, but it was classified as
a suicide attempt, of course.
More overdose examples. You see in this
first example there were 113 tablets and it
exemplifies how medication types were blinded so
you see all the different numbers there. Then, the
next one is patient aged 15, impulsively slit her
wrist following an altercation with her mother.
Finally, age 17, attempted suicide by taking 8
tablets after a fight with her father, whom she
considered harsh and rejecting.
Now, these are examples of self-injurious
behavior, intent unknown. So, this is where we did
some harm but we just don't know why. A patient
aged 10 had superficial scratches, left arm,
scratched herself with scissors. That was all the
information that was there essentially.
Patient, aged 14, ingested or
ingestion of 2-3 cigarettes. The patient was
reported as feeling tired and playing a theatrical
role. Subject, aged 9, reported he had ingested 4
of his brother's tablets on a dare. Finally,
patient, aged 10, swallowed a small amount of
after-shave lotion while angry. It is hard to know
what to make of those without information.
Examples of preparatory actions, age 16,
tried to hand herself and was prevented from doing
so by her family. The next case, age 18, a voice
commanded him to jump from the roof. Although he
went up, he did not jump. Next one, age 10, held a
kitchen knife to her neck while alone but did not
cut herself. Event was not witnessed. Finally, a
patient, age 18, was noted to be hostile, hopeless
and helpless and had written suicide notes. As I
said, anything beyond a verbalization was
considered a preparatory action, including writing
a suicide note.
These are good examples of self-injurious
behavior, no suicidal intent. In the first case
the patient stated there was increased
tension. She made superficial cuts on her wrist
with an Exacto knife. The patient and mother
reported the cuts weren't deep and they looked like
cat scratches. Patient adamantly denied any
suicidal gestures or intent. She stated she only
wanted a release and that cutting and hitting her
legs offers her a release.
The second case, denied suicidal thoughts.
The first time she cut herself was age 16. She
stated she did it for attention. Today her cutting
was more spontaneous. She reported that cutting
gives her a good weird feeling.
So, what were the results? This slide
just refers to what we are talking to in our
results, or referring to, and there were 427 events
but some patients had more than one event so we
ultimately, in the reliability data, used 378
cases. We employed the same severity hierarchy
that the FDA used. So, we just took the most
severe event for cases that had multiple events.
Expert rater consensus--only two of 427
cases had no agreement among the three
each rater had a different rating in only two
cases. Fifty-nine cases had agreement among two of
three raters, and those had to go to
teleconference. There were no cases in which
consensus was not able to be reached during the
teleconference and they, of course, had that
Now, discordant cases between the sponsor
and Columbia classifications, there were 40 out of
the 427 cases in which the sponsor and the Columbia
classification differed. Twenty-six new cases were
identified that had not been identified by the
sponsor as possibly suicide-related. There were
two new cases of self-injurious behavior without
suicidal intent that had been labeled something
other than deliberate self-harm, and 12 cases were
originally called possibly suicidal and were
changed to something other than possibly suicidal.
Here it breaks it down for you further.
Of the 26 new possibly suicide-related events, one
was a suicide attempt; one was a preparatory act;
13 were ideation events; four were intent
acts; and seven were not enough information to say
whether there was deliberate self-harm.
Here is an example of one of the newly
identified suicidal events. This is a preparatory
act. The patient, age 11, held a knife to his
wrist and threatened to harm himself. The patient
was hospitalized with an acute exacerbation of
major depressive disorder. The reason we have this
is because, as I mentioned before, every
hospitalization is a serious adverse event so that
is why this preparatory act was caught.
The events that were changed from suicidal
to something other included two changed to
psychiatric; one changed to an accident; and nine
changed to self-injurious behavior with no suicidal
intent. Again, our famous example, a patient
reported to have engaged in an episode of
auto-mutilation where she slapped herself in the
face. The event resolved the same day without any
These are actually the kappa, the
agreement between the sponsor and
Columbia's classification of possibly
suicide-related and the sponsor's classification of
possibly suicide-related, the kappa was 0.77. You
see in the 2 X 2 table that the numbers correspond
to the numbers that I just went through with you.
Now, if you want to look somewhat more
specifically or at least what we think is more
specific, we did a comparison of what Columbia said
was definitively suicidal and what the sponsor said
was possibly suicidal, and the kappa was 0.69.
Here are the reliability results of the
ratings with the nine expert raters. The median
ICC was 0.86 and what the FDA will refer to as the
primary outcome variables, you can see the numbers
here, suicide attempts is 0.81; preparatory
actions, 0.89; suicidal ideation, 0.97.
Where do we go from here? We need to
improve our adverse event reporting for
suicide-related events by developing consistent
terminology; developing guidelines for
classification of suicidality so that adequate
information is provided by the clinician;
utilization of research assessment tools, what
questions to ask, how to ask, and what measures aid
this; finally, hopefully, that will lead to
improved, more valid identification and
documentation of suicidality.
DR. GOODMAN: Thank you very much. Dr.
DR. CHESNEY: Thank you. This is a little
bit of thinking outside the box, but we heard at
the February meeting a number of examples of
homicidal behavior. I wonder if, in your
speciality, homicidal behavior is ever identified
as being self-injurious primarily to affect
circumstance or to affect an internal state.
DR. POSNER: No, that did not represent
any of those self-injurious, no suicidal intent
ratings. So, the classifications that you are
referring to, internal state and circumstance, are
not synonymous at all with the cases that had
homicidal ideation or any kind of aggressive
behavior. It doesn't mean that it couldn't be
looked at in another analysis but it
represented in these cases.
DR. CHESNEY: I guess my more general
question, I just wonder in the bigger question,
homicidal behavior outcome is bound to be bad and
self-injurious, and if it is just another factor
that we should consider in this whole picture.
DR. GOODMAN: Dr. Robinson?
DR. ROBINSON: Do you know how many of the
events led to hospitalization and how it breaks
down in terms of your classification?
DR. POSNER: Dr. Laughren, do you know?
DR. LAUGHREN: I don't have that figure
off the top of my head. There were a substantial
number of events leading to hospitalization, I
believe somewhere in the ballpark of maybe 40
percent. I don't have the exact number. It was a
DR. POSNER: It is important to know that
we were very narrowly just looking at obtaining the
most appropriate label for the particular event in
question, and we didn't have any of the
information or follow-up information in this
particular piece of the project.
DR. GOODMAN: Dr. Wang?
DR. WANG: I have a question. Do you know
how many of the sponsors originally submitted
reports that were categorized as serious? The
reason I am asking is to get a sense of how many
cases may not have been originally reported. I
know you had these serious cases sent to you for
adjudication, just to check in case there were
cases that were being missed in what the sponsors
were reporting, but did you look as to how many
cases were not considered serious by the sponsor,
jut to give us a sense of how many may be sort of
out there in the non-serious pool?
DR. POSNER: Again, we were blinded to
sponsors' classifications throughout the entire
process. I don't know if somebody from the FDA can
answer that question for you.
DR. LAUGHREN: Again just a ballpark
figure, I think it is probably somewhere in the
vicinity of maybe 65-70 percent. But you have to
understand that a designation of serious is a
judgment that is made by the sponsor fairly
subjectively. I mean, there are criteria for
regulatory serious. It is fatal, life-threatening,
seriously disabling, leading to inpatient
hospitalization. But even though, you know, that
on face appears to be fairly definitive, sponsors
in many cases, in my view, made the judgment that
if it was considered to be suicide-related it was,
by definition, serious.
So, if you look at many of the narratives
that were classified as serious, I think no
reasonable person looking at those would consider
that, in a common sense notion, as a serious event.
But the point is that that designation--how that
judgment was made varied from sponsor to sponsor.
So, you know, some of them classified many more of
the events as serious than other sponsors. But the
answer to the question is that overall roughly
two-thirds of these events that were included in
the analysis were designated as regulatory serious.
DR. GOODMAN: Ms. Griffith?
MS. GRIFFITH: I have a question about
cutting specifically. It seems to me that most of
the examples of cutting fall into self-injurious
behavior, intent unknown or self-injurious behavior
with no intent. I am just curious as to are you
confident that the reporting that you received and
reviewed actually got to whether or not there was
intent or no intent, and how subjective is the
reporting likely to be?
DR. POSNER: Again, as you can see,
cutting is a method that is used both in suicidal
behavior and self-injurious behavior without
suicidal intent. If you remember the conceptual
scheme, there was the category self-injurious
behavior, intent unknown, because cutting can be
used both ways. I forget the exact number but
there were 20-something cases in which they cut but
we don't know if it was suicidal or not. That is
why we had to come up with a category just to
categorize and deal with that issue. The FDA will
point out that the included that in the sensitivity
analysis so just in case all of those
suicide-related events, they have those numbers.
DR. LAUGHREN: If the question you are
asking were the narratives lacking in detail, they
absolutely were. These were not by any sense
complete descriptions. Ideally, many more
questions would have been asked when these events
occurred to help flesh them out. That is why it
was necessary to use inference as one approach to
try and get at intent because intent was not
included for the vast majority of these.
DR. POSNER: Which is why only experts in
the field could have been able to infer from the
surrounding information. The narratives were
limited with respect to suicidal intent often but