1
DEPARTMENT OF HEALTH AND HUMAN SERVICES
FOOD AND DRUG ADMINISTRATION
CENTER FOR DRUG EVALUATION AND RESEARCH
PSYCHOPHARMACOLOGICAL DRUGS ADVISORY COMMITTEE
NDA 21-431: ACAMPROSATE 333 MG TABLETS
LIPHA PHARMACEUTICALS, INC.
Friday, May 10, 2002
8:00 a.m.
Holiday Inn Bethesda
Versailles I
8120 Wisconsin Avenue
Bethesda, Maryland
2
PARTICIPANTS
Dan Oren, M.D., Acting Chairperson
Sandra Titus, Ph.D., Executive Secretary
MEMBERS
Edwin H. Cook, Jr., M.D.
Richard P. Malone, M.D.
Irene E. Ortiz, M.D.
Matthew V. Rudorfer, M.D.
Andrew Winokur, M.D., Ph.D.
STATISTICS SGES
Robert Hamer, Ph.D.
Andrew C. Leon, Ph.D.
PSYCHIATRIC SGE
Paul Keck, M.D.
ALCOHOLISM SGE
Richard Fuller, M.D.
GUESTS (Non-Voting)
John R. Hughes, M.D.
Charles O'Brien, M.D., Ph.D.
Linda Porrino, Ph.D.
Alan Schatzberg, M.D.
INDUSTRY GUEST (Non-Voting)
Dilip J. Mehta, M.D., Ph.D.
FDA
Sandra L. Kweder, M.D.
Cynthia McCormick, M.D
Celia Winchell, M.D.
Sue Jane Wang, Ph.D.
3
C O N T E N T S
Call to Order, Introductions:
Dan Oren, M.D., Acting Chair 4
Conflict of Interest Statement:
Sandra Titus, Ph.D. 7
Welcome and Introductory Comments:
Cynthia McCormick, M.D. 11
Lipha Presentations
Introduction: Anita M. Goodman, M.D. 17
European Development Program and Current
Registration Status:
Silvie Chabac, M.D. 26
Acamprosate: Mechanism of Action, Preclinical
Effects and Pharmacokinetic Overview:
George F. Koob, Ph.D. 29
Efficacy Results from three Pivotal
Clinical Trials:
Karl F. Mann, M.D. 40
Analysis of the U.S. Study Results:
Barbara J. Mason, Ph.D. 50
Closing Remarks:
Anita M. Goodman, M.D. 77
Questions from the Committee to Lipha 82
FDA Presentations
Clinical Issues of Efficacy:
Celia Winchell, M.D. 113
Statistical Perspective of Acamprosate Experience:
Sue Jane Wang, Ph.D. 141
Questions from the Committee to FDA 166
Open Public Hearing
Victor Hesselbrock, Ph.D. 186
Steven Mirin, M.D. 192
Edward Eder, M.D. 197
Mark Publicker, M.D. 199
Charge to the Committee: Cynthia McCormick, M.D. 202
Continuation of Discussion 205
4
1 P R O C E E D I N G S
2 Call to Order, Introductions
3 DR. OREN: Good morning. My name is Dan
4 Oren. I would like to call to order this meeting
5 of the Psychopharmacological Drugs Advisory
6 Committee regarding NDA 21-431 for acamprosate 333
7 milligram tablets.
8 This committee is purely an advisory
9 committee so I have the distinct pleasure of being
10 an acting chair of a committee with no power. But
11 we have an important mission and that mission is to
12 make recommendations to answer questions to give
13 some guidance to the FDA to do with what they wish.
14 I would like the members of the panel to
15 each introduce themselves. We will go around. I
16 will start with the FDA representatives who are
17 from the Review Division and ask--Sandy Kweder is
18 not here yet but we will start with Dr. Cynthia
19 McCormick.
20 DR. McCORMICK: How do you do. I am Dr.
21 Cynthia McCormick. I am the Director of the
22 Division of Anesthetic, Critical Care and Addiction
23 Drug Products at the FDA. Welcome.
24 DR. WINCHELL: I am Celia Winchell. I am
25 the Medical Team Leader for Addiction Drug Products
5
1 and I did the primary clinical review for this NDA
2 DR. WANG: Good morning, everyone. My
3 name is Sue Jane Wang. I am the Statistic Leader
4 in the Alcoholism Treatment Clinical Trials. I am
5 the statistical reviewer for this project.
6 DR. OREN: As we go around to the formal
7 members of our panel, I would introduce Dr. Leon is
8 who is a new member. I want to ask everyone, in
9 addition to telling us a little bit about what you
10 do, tell us where you are from.
11 DR. LEON: I am Andrew C. Leon, Cornell
12 University Medical College. I work primarily in
13 affective disorders and anxiety disorders.
14 DR. KECK: My name is Paul Keck. I am
15 Vice Chair for Research in the Department of
16 Psychiatry at the University of Cincinnati College
17 of Medicine.
18 DR. HAMER: I am Bob Hamer. I am
19 Professor of Psychiatry and Biostatistics at the
20 University of North Carolina.
21 DR. WINOKUR: Andy Winokur from the
22 Department of Psychiatry, University of Connecticut
23 Health Center. I am Director of Psychopharmacology
24 there.
25 DR. MALONE: I am Richard Malone from the
6
1 Department of Psychiatry at the Medical College of
2 Pennsylvania in Philadelphia. I am involved mainly
3 in child psychiatry research.
4 DR. RUDORFER: I am Matthew Rudorfer from
5 the National Institute of Mental Health. I am the
6 Associate Director for Treatment Research in the
7 Division of Services and Interventions Research.
8 DR. TITUS: I am Sandy Titus. I am with
9 the FDA. I am the Executive Secretary for PDAC.
10 DR. OREN: I am still Dan Oren. I am an
11 Associate Professor of Psychiatry at Yale
12 University.
13 DR. ORTIZ: I am Irene Ortiz. I am from
14 the Department of Psychiatry at the University of
15 New Mexico and the Albuquerque V.A. I am in
16 geriatric psychiatry and addiction psychiatry.
17 DR. FULLER: I am Richard Fuller. I am
18 Director of the Division of Clinical and Prevention
19 Research at the National Institute on Alcohol Abuse
20 and Alcoholism.
21 DR. PORRINO: I am Linda Porrino, a
22 Professor in the Department of Physiology and
23 Pharmacology at Wake Forest University School of
24 Medicine
25 DR. HUGHES: I am John Hughes. I am a
7
1 Professor in Psychiatry at the University of
2 Vermont.
3 DR. MEHTA: I am Dilip Mehta. I am the
4 industry representative on this committee.
5 DR. OREN: Drs. Porrino, Hughes and Mehta
6 are guests with the committee and we are delighted
7 to have you here.
8 The three questions that we have before us
9 for today are we are asked to consider the evidence
10 of efficacy of acamprosate in the treatment of
11 alcoholism and to provide advice on three key
12 questions.
13 One, how can the discrepant results
14 between the older European studies and the more
15 recently conducted American study be reconciled?
16 Two, do the data support any conclusions regarding
17 subgroups of patients more likely to benefit from
18 acamprosate? Three, given the conflicting results,
19 is there sufficient evidence of the efficacy of
20 acamprosate in the treatment of alcoholism to
21 warrant approval?
22 I will turn the podium over to Dr. Titus.
23 Conflict of Interest Statement
24 DR. TITUS: I am going to read the
25 conflict of interest statement dealing with
8
1 acamprosate for this meeting. The following
2 announcement addresses the issue of conflict of
3 interest issues associated with this meeting and is
4 made a part of the record to preclude even the
5 appearance of such at this meeting.
6 Based on the submitted agenda for the
7 meeting and all relevant financial interests
8 reported by the committee participants, it has been
9 determined that all interest in firms regulated by
10 the Center for Drug Evaluation and Research present
11 no potential for an appearance of a conflict of
12 interest with the following exceptions.
13 Robert Hamer has been granted waivers
14 under 18 USC 208(b)(3) and 21 USC 355(n)(4) for his
15 and his spouse's stock in the parent company of a
16 competitor. The stock is valued between $25,000 to
17 $50,000.
18 Richard Fuller has been granted a waiver
19 under 21 USC 355(n)(4) for his stock in the parent
20 company of a competitor. The stock is valued from
21 $5,001 to $25,000. Because 5 CFR 2640.202(a) de
22 minimis exemption applies, a waiver under 18 USC
23 208(b)(3) is not required.
24 Paul Keck has been granted a waiver under
25 21 USC 355(n)(4) for his stock in the parent
9
1 company of a competitor. The stock is valued at
2 less than $5,001. Because 5 CFR 2640.202(a) de
3 minimis exemption applies, a waiver under 18 USC
4 208(b)(3) is not required.
5 A copy of the waiver statements may be
6 obtained by submitting a written request to the
7 Agency's Freedom of Information Office, Room 12A30
8 of the Parklawn Building. With respect to FDA's
9 invited guests, there are reported interests that
10 we believe should be made public to allow the
11 participants to objectively evaluate their
12 comments. Dr. Anthony Schatzberg consulted with
13 Bristol-Myers Squibb within the past year on a drug
14 which is unrelated to acamprosate or its competing
15 products.
16 Dr. Charles O'Brien is Chief of Psychiatry
17 at the Philadelphia Veterans Affairs Medical
18 Center. Dr. O'Brien was previously invited by
19 Forest Laboratories to a meeting concerning
20 acamprosate. However, he was unable to attend due
21 to other commitments. Dr. O'Brien was the first to
22 initiate a study of naltrexone, a competing product
23 in alcoholism and his center participated in the
24 U.S. acamprosate trial. But he had no direct
25 involvement. His center is also participating in
10
1 the NIH-sponsored study of naltrexone and
2 acamprosate but he is not directly involved. Dr.
3 O'Brien previously received consultant and speaker
4 fees from Dupont Pharmaceuticals. Lastly, he has
5 been invited to be a member of Forest Laboratories
6 Advisory Board but he had not yet accepted.
7 In addition, we would like to disclose
8 that Dr. Dilip Mehta is participating in this
9 meeting as an industry guest acting on behalf of
10 regulated industry. Dr. Mehta reported that he
11 owns stock in Bristol-Myers Squibb.
12 In the event the discussions involve any
13 other products or firms not already on the agenda
14 for which an FDA participant has a financial
15 interest, the participants are aware of the need to
16 exclude themselves from such involvement and their
17 exclusion will be noted for the record.
18 With respect to all other participants, we
19 ask, in the interest of fairness, that they
20 address any current or previous financial
21 involvements with any firm whose products they may
22 wish to comment upon.
23 Thank you.
24 DR. OREN: I will now call upon Dr.
25 Cynthia McCormick, Director of the Anesthetic
11
1 Critical Care and Addiction Drug Products at the
2 FDA.
3 Welcome
4 DR. McCORMICK: Thank you. Dr. Chairman,
5 Advisory Committee Members, Invited Guests, members
6 of FDA and members of public, welcome to this
7 meeting of the Psychopharmacologic Drugs Advisory
8 Committee convened to discuss the efficacy of
9 acamprosate.
10 Chronic alcoholism continues to be a
11 widespread and debilitating disorder which places a
12 tremendous burden on society in healthcare costs,
13 lost wages and personal suffering. The need for
14 effective pharmacologic agents for this disorder
15 cannot be overstated. It has been estimated that
16 100,000 lives and $184.6 billion annually are the
17 cost of chronic alcoholism in the United States.
18 Currently, there are only two
19 pharmacologic agents available for alcoholism in
20 the U.S. Antabuse was approved in 1951 and
21 marketed at times intermittently. Revia,
22 containing the opioid antagonist naltrexone, was
23 approved for this indication in 1994.
24 Despite the crying need for new and better
25 pharmacotherapies, it is very important that drugs
12
1 approved for this condition must meet the FDA
2 standards for safety and effectiveness. To approve
3 a drug with marginal effectiveness or no
4 effectiveness at all would have no more
5 public-health benefit than to approve no drug.
6 In December, 2001, the FDA received for
7 review a new drug application for the product
8 acamprosate. Acamprosate has been available in
9 Europe for the treatment of chronic alcoholism for
10 nearly fifteen years. The application, when filed,
11 was given a priority review status by the FDA
12 because this was hoped to have the potential to
13 affect the course of a disease with tremendous
14 morbidity and mortality.
15 The FDA team has completed the review of
16 the efficacy of this product and has struggled with
17 the contradictory efficacy results between the
18 European and United States study. The efficacy
19 data on which this application rests includes a
20 number of European clinical trials performed over
21 the last fifteen years, three of which are
22 considered pivotal studies, and a recently
23 completed U.S. multicenter trial.
24 The results of these studies on their face
25 paint a conflicting picture. The FDA team has
13
1 attempted to explore the apparent contradictions by
2 evaluating the differences between these studies
3 through a variety of analyses. The discussion of
4 these factors and how they contribute to our
5 understanding of the drug's efficacy will be the
6 primary focus of this meeting.
7 The three pivotal European trials, Pelc
8 II, Paille and PRAMA were of similar design,
9 methodology and outcomes. The trials have been
10 considered successful by the company and the review
11 team concurs with this assessment but with caveats
12 which the FDA team will be reviewing this morning.
13 The U.S. study, on the other hand, was not
14 successful in demonstrating superiority over
15 placebo on the primary outcome and most secondary
16 measures. Indeed, on some measures, the drug
17 appeared to perform less well than placebo.
18 Some differences between the European and
19 U.S. studies can be clearly delineated. The
20 European population was primarily one of pure
21 alcoholics. The U.S. population was largely
22 polysubstance abusers. The European patients had
23 either recently undergone detoxification and were
24 abstinent prior to randomization. The U.S.
25 patients were generally not abstinent prior to
14
1 randomization.
2 The ascertainment of drinking data in the
3 European studies was essentially retrospective,
4 infrequent and the values were heavily imputed. It
5 was very methodical and rigorous in the U.S. study
6 using accepted methods for reconstructing drinking
7 data and information was obtained at frequent
8 intervals. There were also very tight follow-up
9 provisions in place in the U.S. study.
10 The review team has attempted to apply the
11 same conservative approach to analysis of the data
12 of the U.S. and the European studies but they have
13 obtained disparate results.
14 Finally, the studies differed in terms of
15 the formulation of acamprosate that was used and
16 the regimen of administration, although the total
17 daily dose was essentially the same.
18 It is not uncommon for an NDA database to
19 have both successful results and results which are
20 not considered positive. In general, the agency's
21 approach to such a situation is to consider the
22 totality of the evidence giving consideration and
23 weight to such factors as the quality of the data,
24 the strength of the effect size, statistical
25 significance and assessment of whether the effects,
15
1 even in the negative trials, are supportive or
2 trend in the right direction and are not
3 contradictory.
4 If a trial has truly failed--that is,
5 demonstrated an effect that contradicts the
6 remainder of the evidence--an attempt is made to
7 understand the reason for that contradiction and to
8 determine, on balance, which results are more
9 credible. Occasionally, further clinical work is
10 needed.
11 In this NDA, the differences between the
12 studies are clear. The questions that remain,
13 however, are whether these differences can
14 adequately account for the disparate results and
15 whether the failure of acamprosate in the U.S.
16 study was a function of the difference in
17 responsiveness of the U.S. alcoholic population or,
18 perhaps, a difference in manifestation of the
19 disease.
20 Stated differently, can the results of the
21 European trials be generalized to the U.S.
22 alcoholic population? There are other aspects of
23 the drug-approval decision which are not being
24 brought for discussion today. The drug's safety is
25 still under evaluation and is expected to be
16
1 completed at the end of this month.
2 Both clinical inspections and inspections
3 of the manufacturing site have not been conducted
4 and are expected to be conducted by the end of
5 June. These will both be weighed into the decision
6 for approval and also in the timing of approval.
7 For this reason, the advisory committee meeting
8 today will not be one in which a final approval
9 recommendation is being requested.
10 The FDA is seeking the advice of the
11 Psychopharmacologic Drugs Advisory Committee and
12 experts in clinical research in alcoholism on your
13 assessment of the evidence provided in support of
14 the efficacy of this product. We are inviting the
15 committee to discuss a series of questions probing
16 the issues surrounding the efficacy results and to
17 make recommendations that will ultimately aid the
18 FDA in making its determination once the other
19 aspects of the application are complete.
20 These will lead to the final decision
21 about the approvability of the product for the
22 maintenance of abstinence in chronic alcoholism.
23 Thank you.
24 DR. OREN: Thank you, Dr. McCormick.
25 I would like to ask three more members of
17
1 our panel who arrived to introduce yourselves, tell
2 us who you are, where you are from.
3 Dr. Cook?
4 DR. COOK: Dr. Cook, University of
5 Chicago.
6 DR. OREN: Dr. Schatzberg?
7 DR. SCHATZBERG: Dr. Schatzberg from
8 Stanford University.
9 DR. OREN: Dr. O'Brien?
10 DR. O'BRIEN: Charles O'Brien, University
11 of Pennsylvania.
12 DR. OREN: We will now move on to the
13 presentations by Lipha. I would like to introduce
14 Dr. Anita M. Goodman, Executive Vice President and
15 Chief Operating Officer of Lipha.
16 Lipha Presentations
17 Introduction
18 DR. GOODMAN: Good morning.
19 [Slide.]
20 I am Anita Goodman of Lipha
21 Pharmaceuticals. I would like to introduce our
22 presentation on acamprosate, a new therapy for
23 maintaining abstinence in alcohol dependence.
24 [Slide.]
25 Alcohol dependence is a medical disorder
18
1 which afflicts at least 8 million Americans with
2 almost an equal number of alcohol abusers. The
3 cost to society of alcohol dependence are enormous,
4 both in terms of medical and hospitalization costs,
5 losses and economic potential from reduced
6 productivity and premature death, and costs related
7 to incarceration and judicial process.
8 Beyond the obvious economic implications,
9 costs cannot be attributed to the significant
10 emotional toll this disorder extracts from families
11 affected by alcohol dependence and for the loss of
12 lives, both the lives of patients often still in
13 their prime and the innocent lives of those killed
14 by drunk drivers.
15 Every person in this room knows and has
16 been touched by at least one person whose entire
17 life has been altered and all too often ruined by
18 alcohol dependence. The effect of that dependence
19 reaches out and extends to everyone who that loves
20 them, that cares about them and that works
21 alongside them.
22 The treatment of alcohol dependence is not
23 easy nor, in its current status, is it uniformly
24 successful. It requires the voluntary engagement
25 and time commitment of the patient, involvement of
19
1 family members and concerned friends, and a team
2 approach of care providers ranging from self-help
3 groups, social workers, to psychologists,
4 psychiatrists and internists.
5 [Slide.]
6 The role of pharmacotherapy in this
7 disorder continues to be limited by lack of
8 available approved medications and possibly also by
9 some resistance on the part of the treatment
10 community to consider alcohol dependence as
11 amenable to treatment by anything except intensive
12 behavioral and psychotherapy.
13 Furthermore, from a product-development
14 point of view, there is the additional lack of
15 universally accepted outcome parameters. This is
16 undoubtedly one of the areas we will touch on
17 today. In contract to diabetes or hypertension,
18 for example, where there are universally agreed to
19 and reliably measurable endpoints for the
20 regulatory assessment of the product's
21 efficacy--for example, hemoglobin A1C and blood
22 glucose levels in diabetes, or blood-pressure
23 changes in hypertension--the ideal parameter or
24 parameters for measuring outcome in trials or
25 therapies for alcohol dependence have not been
20
1 agreed to and, in fact, are still under active
2 discussion both by academicians as well as
3 regulatory agencies.
4 In the studies described in your briefing
5 documents, we have proposed and utilized a group of
6 related parameters linked to self-reported drinking
7 behavior but in the context of a platform of
8 abstinence. The FDA has expressed some concerns
9 about the methodologies in obtaining
10 outcome-related information and we will address
11 this today.
12 Unlike almost every other medical
13 disorder, as Dr. McCormick pointed out, there are
14 only two pharmacotherapies available which are
15 specific for alcohol dependence post-withdrawal,
16 the aversive agent disulfiram and the opioid
17 antagonist naltrexone. Both of these drugs,
18 however, have limitations in their general
19 applicability related to their mechanism of action
20 either because patients may have significant
21 hepatic dysfunction or may slip to drinking.
22 Thus, a medication such as acamprosate
23 that is more encompassing should be welcomed by the
24 treatment community.
25 [Slide.]
21
1 This morning, you will be hearing about
2 the development of acamprosate and its current
3 global registration status from my French
4 colleague, Dr. Silvie Chabac, who is based at
5 Lipha's headquarters in Lyon, France. Dr. Chabac
6 has been involved with acamprosate clinical
7 research for many years and brings considerable
8 knowledge and experience to today's meeting.
9 In her presentation, Dr. Chabac will
10 describe to you the core acamprosate studies which
11 comprise the registration dossier for this product
12 worldwide.
13 [Slide.]
14 Fourteen double-blind placebo-controlled
15 studies were conducted throughout Europe between
16 1989 and 1995. From these studies, thirteen of
17 which supported the efficacy and safety of
18 acamprosate in maintaining abstinence and only one
19 of which showed no significant treatment effect, we
20 selected three as pivotal for the following
21 reasons.
22 [Slide.]
23 All fourteen studies in the clinical
24 portion of the European dossier were conducted by
25 qualified experts who are alcohol specialists
22
1 working in specialized centers or departments. All
2 the studies were performed according to existent
3 standards of good clinical practice. They all
4 followed specific protocols and have existing
5 retrievable case-report forms as well as electronic
6 databases.
7 However, the three studies considered by
8 Lipha to further qualify as pivotal, had the
9 following additional characteristics. The study
10 centers were still active and the source documents
11 and other medical records were still largely
12 accessible, thereby permitting an on-site FDA audit
13 as is required for a new drug application.
14 You have to keep in mind that collectively
15 the archival requirements for retaining documents
16 had been exceeded for the majority of these
17 European studies.
18 [Slide.]
19 In addition, for these three studies, the
20 clinical research organizations or CROs which had
21 managed the trials were still active and also had
22 some of the original trial management
23 documentation. The final point would be that two
24 of these studies looked at two dose levels of
25 acamprosate and they also, thereby, provide some
23
1 suggestion of dose responsiveness.
2 [Slide.]
3 Following Dr. Chabac will be a
4 presentation by Dr. George Koob, Professor and
5 Director of the Neuropharmacology Division of the
6 Scripps Research Institute in La Jolla. Dr. Koob
7 is the recipient of this year's Distinguished
8 Investigator's Award of the Research Society on
9 Alcoholism and the recipient of this year's award
10 from the American Society on Addiction Medicine.
11 Dr. Koob has worked in the area of animal
12 models and mechanisms of alcohol dependence for
13 many years and has provided significant insight
14 into the way in which acamprosate exerts its
15 activity. He is the author of more than 500
16 peer-reviewed articles largely on addiction.
17 This morning, Dr. Koob will discuss
18 acamprosate's preclinical effects, its purported
19 mechanism of action and will also briefly cover
20 acamprosate's pharmacokinetic profile.
21 [Slide.]
22 Dr. Karl Mann, Professor and Chairman of
23 the Department of Addictive Behavior and Addiction
24 Medicine at the University of Heidelberg in Germany
25 and also an investigator in one of the pivotal
24
1 studies will then review for you the efficacy
2 results from the three pivotal European clinical
3 trials.
4 Dr. Mann has the unique distinction of
5 holding the only Chair of Addiction in Germany. He
6 is the European editor of the journal Alcoholism,
7 Clinical and Experimental Research and is the
8 author of more than 200 scholarly papers in the
9 area of clinical research on alcoholism.
10 Because, as Dr. McCormick has mentioned to
11 you, the safety review of acamprosate is still
12 ongoing, we cannot present safety data today, so
13 Dr. Mann's focus will be on efficacy only. As you
14 know from the documents you have received, the FDA
15 has convened this committee and its invited experts
16 to consider the persuasiveness of the data from
17 these European studies for the proposed indication.
18 I would like to point out that we, Lipha,
19 always intended to rely heavily on the substantial
20 European database for this new drug application in
21 our overall development strategy. But we also felt
22 that it was very important to conduct a safety and
23 efficacy study of acamprosate in alcohol-dependent
24 patients in the United States.
25 At the recommendation of the division,
25
1 very broad admission criteria were used. The
2 results of the American trial called U.S. 96.1 in
3 your document which may, at first, appear to be at
4 odds with the conclusions of the European studies
5 which Dr. Mann will describe has, however, afforded
6 us the opportunity to gain further insight into how
7 acamprosate works best.
8 [Slide.]
9 Dr. Barbara Mason, Professor of Psychiatry
10 and Behavioral Sciences and Director of the
11 Substance Abuse Division at the University of Miami
12 as well as overall principle investigator for the
13 American study, U.S. 96.1, will present data from
14 the study and the understanding that it has
15 brought. Again, at the division's request, the
16 discussion will focus on efficacy.
17 Dr. Mason has extensive experience in
18 clinical alcohol research and has been the
19 principle investigator of many NIH-funded clinical
20 trials involving medication development in
21 alcoholism. She serves on the Scientific Advisory
22 Council of the National Institute on Alcohol Abuse
23 and Alcoholism and is field editor for the Journal
24 of Neuropsychopharmacology.
25 Dr. Mason has published extensively
26
1 including in such journals as the Archives of
2 General Psychiatry and the Journal of the American
3 Medical Association.
4 [Slide.]
5 Following her presentation, I will then
6 make some concluding comments and answer or
7 redirect questions you may have.
8 [Slide.]
9 As Dr. McCormick has enumerated, we have
10 been asked by FDA to address the following issues.
11 Why were the efficacy results for the ITT
12 population in the U.S. trial inconclusive in
13 contrast to the consistently positive European
14 studies? Were the methodologies appropriate and
15 are European and American alcohol-dependent
16 populations comparable?
17 In the next hour, our presentations will
18 bring clarity to these issues.
19 Thank you. Dr. Chabac will now speak.
20 European Development Program
21 and Current Registration Studies
22 DR. CHABAC: Good morning.
23 [Slide.]
24 My name is Silvie Chabac. I was the
25 doctor responsible for the European Development of
27
1 Program of acamprosate. I would like to give you
2 an overview of these programs along with the
3 current registration status of acamprosate.
4 [Slide.]
5 The story of acamprosate began in France
6 in the early 1980s. The French pharmaceutical
7 company, Laboratoires Meram, decided to investigate
8 amino-acid neuromediators as a new research
9 project. During the screening tests, one compound
10 was particularly noted for its outstanding
11 pharmacological properties, calcium acetyl
12 homotaurine, now best known as acamprosate.
13 Based on animal work that Dr. Koob will be
14 describing shortly, Meram then decided to
15 specifically develop this compound for alcohol
16 dependence. In 1987, the 333 milligram acamprosate
17 tablet was authorized for marketing authorization
18 in France. It has been commercially available
19 there since 1989.
20 At that stage, Meram transferred the
21 license for acamprosate to its sister company,
22 Lipha, for worldwide development.
23 [Slide.]
24 The same year, Lipha began an extensive
25 clinical program throughout European for
28
1 registration purposes. Over 4000 alcohol-dependent
2 patients were randomized in fourteen double-blind
3 placebo-controlled studies conducted in ten
4 different European countries. This
5 clinical-development program included long-term
6 studies, two phase II and twelve phase III, with
7 the treatment period ranging from 3 to 12 months.
8 [Slide.]
9 Based on the solid efficacy and safety
10 results of this development program, Lipha began
11 the worldwide registration of acamprosate. Today,
12 it is registered in 30 countries on five continents
13 where alcohol dependence is recognized as a disease
14 and a major public-health problem. Since 1995,
15 worldwide registration has been ongoing, first in
16 Europe where it is now approved for marketing in
17 nineteen countries including Scandinavian countries
18 and Eastern Europe, then, in South and Central
19 America and in Mexico. Three years ago,
20 acamprosate was registered in Australia, Singapore
21 and Hong Kong, then last year in South Africa.
22 Finally, to complete the process, Lipha
23 submitted an NDA for acamprosate in the United
24 States of America last December. In every country
25 where it is marketed, acamprosate has a specific
29
1 labeling; maintenance of long-term abstinence in
2 patients with alcohol dependence who have been
3 withdrawn from alcohol.
4 Acamprosate should be prescribed in
5 conjunction with counseling for a recommendation
6 treatment duration of one year. To date, around
7 the world, there have been 1.5 million patient
8 years of exposure. This group of patients had the
9 opportunity to benefit from the treatment with
10 acamprosate for alcohol dependence.
11 Now, we would like to make that
12 opportunity available to patients in the United
13 States.
14 Thank you very much.
15 Acamprosate: Mechanism of Action,
16 Preclinical Effects and Pharmacokinetics
17 DR. KOOB: Good morning.
18 [Slide.]
19 I am George Koob. I have been consulting
20 with Lipha Pharmaceuticals since 1990,
21 approximately eleven or twelve years, on their
22 preclinical program.
23 [Slide.]
24 Acamprosate or calcium acetyl homotaurine
25 is the calcium salt of acetylated homotaurine.
30
1 Homotaurine is a homolog of the naturally occurring
2 amino acid taurine and does not readily cross the
3 blood-brain barrier. The acetylation of
4 homotaurine makes the compound more lipophilic and
5 allows penetration of the blood-brain barrier by
6 this compound.
7 I am going to discuss with you, very
8 briefly, this morning the neuropharmacological
9 mechanism of action of acamprosate, its
10 pharmacokinetics and its interactions with other
11 drugs or, shall I say, its lack of interaction with
12 other drugs.
13 [Slide.]
14 The neuropharmacological mechanism of
15 action of acamprosate has been elucidated by
16 extensive use of animal models. Animal models of
17 alcohol have evolved significantly over the past
18 twenty years and have a high degree of face and
19 predictive validity.
20 [Slide.]
21 The animal models for understanding the
22 actions of acamprosate can be understood in terms
23 of excessive drinking, excessive drinking that is
24 driven by dependence, abstinence and relapse. I am
25 going to give you one clear example of the actions
31
1 of acamprosate preclinically in an animal model of
2 excessive drinking.
3 [Slide.]
4 Before I do that, let me just review with
5 you very quickly the evidence that was accumulating
6 on acamprosate through animal models. It is
7 critical for you to understand that, in all of
8 these models, the animals were producing an
9 excessive amount of alcohol intake by a variety of
10 means. Acamprosate decreases alcohol drinking in
11 rats that were selected for excessive drinking. In
12 one of the earliest studies, acamprosate decreases
13 alcohol intake in dependent animals. This is
14 another one of the early studies done by Le Magnin
15 group. Acamprosate reverses the preference for
16 alcohol and the increase in drinking in dependent
17 animals during withdrawal.
18 I am going to show you an example from our
19 own laboratory where acamprosate eliminates the
20 alcohol deprivation effect in rats under
21 free-drinking operant limited-access conditions.
22 [Slide.]
23 Rodents, like human beings, and this
24 speaks to the face validity of the animal models,
25 don't like the taste of alcohol so, to induce a
32
1 rodent to drink alcohol, one starts with a sweet
2 solution. We, in our laboratory, use saccharine,
3 and fade in alcohol and ultimately fade out the
4 saccharine. These animals are in limited-access
5 situations where they drink alcohol once in the
6 evening. They have a lever that they press to
7 obtain 10 percent alcohol or water. By the end of
8 a two- or three-week period, these animals are
9 drinking pharmacological amounts of alcohol in this
10 30-minute session.
11 [Slide.]
12 You can see that the alcohol intakes range
13 from about 20 to 80 milligram percent which is
14 equivalent to what you or I would have from one
15 glass of wine. In doing this kind of a procedure,
16 you can reliably have a baseline drinking of
17 alcohol but you can also make manipulations that
18 will produce increases in drinking that at least
19 have face validity and some predictive validity for
20 the human condition.
21 [Slide.]
22 In this side, what I am showing you is if
23 you stop the animal's availability to alcohol for a
24 series of days, what you see is an increase in
25 alcohol intake that is quite dramatic when the
33
1 animal is reinstated. This is called the alcohol
2 deprivation effect in rodent models. It is
3 equivalent to the abstinence violation effect in
4 human alcoholics.
5 What you can see from this slide is that
6 the animals that have three, five, seven or
7 fourteen days of abstinence between their
8 self-administration show a dramatic increase in the
9 amount of alcohol. They show a dramatic increase
10 in their blood-alcohol levels, jumping from
11 30 milligram percent, on average, to approximately
12 80 milligram percent.
13 What you can also see is, that on a
14 baseline condition, the behavior is very stable in
15 this model. You can also see here that there is no
16 effect of the alcohol deprivation effect on water
17 intake.
18 [Slide.]
19 Acamprosate dose-dependently, as in other
20 animal models of excessive drinking, decreases the
21 alcohol deprivation effect. There are a couple of
22 very important points from this slide from the
23 point of view of the animal models and the
24 preclinical effects of acamprosate.
25 One is that acamprosate has no effect on
34
1 baseline drinking at these doses. Higher doses
2 will affect baseline drinking. The other issue is
3 that acamprosate has no effect on water intake.
4 Both of these argue to the selectivity of the
5 effect on excessive drinking and the selectivity of
6 the effect from the point of view of other
7 behavior.
8 [Slide.]
9 It is also important to note what
10 acamprosate does not do in animal models. What
11 this slide addresses is that acamprosate does not
12 produce what we would call anxiolyticlike effects
13 or anticonflict effects in animal models of
14 anxiety. The other points on this slide just
15 simply illustrate the fact that acamprosate has no
16 abuse potential in preclinical animal models.
17 Acamprosate does not substitute for
18 alcohol. It does not block the discriminative
19 stimulus properties of alcohol. It doesn't have
20 any reinforcing or aversive effects on its own and
21 it doesn't interact with other drugs of abuse.
22 [Slide.]
23 The neuropharmacological mechanism of
24 action of acamprosate is thought to focus on three
25 major areas as depicted in this slide. Acamprosate
35
1 is thought to modulate glutamate receptors as
2 illustrated by No. 1 here. Acamprosate is thought
3 to modulate voltage-dependent calcium channels and
4 acamprosate may also have long-term effects on
5 intermediate early gene products that can
6 ultimately change subunit expression of glutamate
7 receptors. Glutamate, as you know, is the major
8 excitatory neurotransmitter in the brain.
9 [Slide.]
10 More specifically, the
11 neuropharmacological effects of acamprosate can be
12 shown in the following way. Acamprosate has been
13 clearly shown to inhibit neuronal hyperexcitability
14 by decreasing presynaptic release of the excitatory
15 neurotransmitter, glutamate, and by decreasing
16 postsynaptic excitability of glutamate receptors.
17 Acamprosate, as I mentioned, inhibits
18 calcium influx through the NMDA glutamate receptor
19 possibly through an interaction with the polyamine
20 site on the NMDA receptor. This is very important
21 for understanding its action because this is a
22 modulatory effect. It is thought to be an
23 allosteric interaction. It is not a direct
24 receptor action. Acamprosate is not MK801, for
25 those of you versed in this. That means that it is
36
1 not a noncompetitive antagonist to the glutamate
2 receptor. It does not interact directly with any
3 receptor component of the glutamate receptor that
4 would lend it to toxicity.
5 Acamprosate also inhibits calcium influx
6 through voltage-dependent calcium channels. Just
7 to add to its profile as an antihyperexcitability
8 agent, acamprosate also increases the synaptic
9 availability of the inhibitory neurotransmitter
10 taurine, the work of Philip De Witte and his
11 colleagues.
12 [Slide.]
13 What does this mean? What it means, very
14 simply, is that acamprosate acts a partial
15 coagonist at the glutamate receptor through an
16 allosteric interaction with the polyamine binding
17 site on the NMDA glutamate receptor complex.
18 What does this translate to? It
19 translates to a normalization of the receptor
20 system that has become disregulated by the chronic
21 administration of alcohol. That statement, itself,
22 has all the key elements there. It is the
23 normalization of a disregulated receptor system and
24 neurotransmitter system that has been disregulated
25 by chronic alcohol and chronic withdrawal and
37
1 repeated alcohol and repeated withdrawal.
2 So neuropharmacological consequences are
3 to enhance activation of the glutamate receptor
4 when endogenous levels of the activators such as
5 glutamate are low but, most critically, to inhibit
6 activation when levels of the endogenous activators
7 are high such as during alcohol withdrawal.
8 [Slide.]
9 This lead to further observations of great
10 scientific interest that acamprosate also has
11 neuroprotective actions. I am not going to go
12 through the details but simply to say that, in a
13 number of in vitro and preclinical models,
14 acamprosate has been shown to have
15 neuroprotective-like effects.
16 [Slide.]
17 From the point of view of the
18 pharmacokinetics, I thought I would go through this
19 and spend a little time on this. Acamprosate does
20 have bioavailability and that has been adequately
21 demonstrated presumably because of the
22 modifications of the molecule to make it more
23 lipophilic. That is 11 percent in humans, 16
24 percent in rodents.
25 It has an elimination half-life of 18
38
1 hours in humans. Similar, a little longer, in
2 rodents. The time-to-steady-state plasma levels is
3 five to seven days, a critical issue in regards to
4 the design of preclinical studies and clinical
5 studies. There is no protein binding of
6 acamprosate in the blood.
7 The most critical point on this slide is
8 that the elimination of acamprosate is by renal
9 excretion. It is not metabolized and, thus,
10 hepatically compromised patients do not have to
11 worry about taking acamprosate.
12 The lethality in humans, there is no known
13 lethality. In rodents, the dose that has produced
14 lethality is 6 grams per kilogram. This is several
15 log units higher than the effective dose. It is
16 way out there.
17 [Slide.]
18 Acamprosate has basically no interactions
19 with any alcohol. Other drugs that are used for
20 the treatment of alcoholism and other
21 psychotherapeutic drugs with the exception that
22 there is data in press--Dr. Mason has a paper in
23 press in Neuropsychopharmacology showing that
24 naltrexone actually increases plasma levels of
25 acamprosate by about 25 to 30 percent depending on
39
1 what measure you are using.
2 The mechanism for that increase in plasma
3 acamprosate levels by naltrexone is not known but
4 probably has something to do with its pericellular
5 mechanism of absorption.
6 [Slide.]
7 So I would like to stop now just with this
8 slide to reiterate the neuropharmacologic mechanism
9 of action of acamprosate. This
10 neuropharmacological action of acamprosate, as I
11 said earlier, has three major components. The
12 bottom line is that acamprosate normalizes the
13 hyperexcitability in the brain associated with
14 alcohol dependence, notably alcohol withdrawal and
15 protracted abstinence.
16 Acamprosate does this by modulating the
17 glutamate receptor as a partial agonist which is a
18 very effective pharmacological way of returning the
19 brain to a normal state. It also modulates
20 voltage-dependent calcium channels and it also
21 interacts with the taurine neurotransmitter which
22 is an inhibitory neurotransmitter also decreasing
23 neuronal hyperexcitability.
24 Thank you.
25 Efficacy Results from Three Pivotal Clinical Trials
40
1 DR. MANN: Good morning. I'm Karl Mann.
2 I am working at the University of Heidelberg. It
3 is my pleasure to share some of the data that we
4 gained about ten years ago in these European trials
5 with acamprosate. Apart from my academic
6 affiliation, I also run a hospital with inpatient
7 and outpatient treatment for alcoholics where we do
8 treat patients with acamprosate on a day-to-day
9 basis so we could share also some of the
10 experiences that we have been gaining there with
11 you today.
12 [Slide.]
13 I am going to talk about these three
14 studies which were done in Europe about ten years
15 ago. Their objective was to look at the safety and
16 efficacy of acamprosate versus placebo in
17 maintaining long-term abstinence in alcoholics
18 following alcohol withdrawal.
19 [Slide.]
20 These studies were done in Belgium, in
21 Germany and in France. They were all multicenter,
22 like twelve centers in this study and twelve
23 centers in the German study, 31 centers in the
24 French study with a large number of patients
25 included which wound up to almost 1,000 patients in
41
1 these three studies.
2 [Slide.]
3 As I said, they were double-blind
4 randomized and placebo-controlled, all three of
5 them. They were multicenter. Two of them used two
6 dosage levels like the Pelc study and the Paille
7 study in France. They had one arm with a medium
8 dose of acamprosate and one arm with about 2 grams
9 of acamprosate per day whereas the German study had
10 only one arm of medication versus placebo.
11 The Pelc study was done over a period of
12 three months, twelve weeks, and there was no
13 after-care after that whereas the other two studies
14 were over a period of a whole year, so a whole year
15 of study. Then, in the German study, another
16 twelve months of investigation of looking how the
17 patients did afterwards. In the Paille study in
18 France, this was six months.
19 It is also important to note that the
20 psychosocial therapy that was provided to the
21 patients was site-specific. There was not one
22 psychosocial treatment for everybody across all
23 sites. But, of course, within the sites, those
24 patients who received acamprosate or placebo also
25 received the same kind of psychosocial treatment.
42
1 [Slide.]
2 We had male and female patients, of
3 course, in these studies. We had a lower and an
4 upper age limit so no patients who were older than
5 65 years were allowed. They were all DSMIII or
6 DSMIII-R, positive alcoholics. They all, and this
7 is another important point, had detoxification
8 prior to the entry into this study.
9 This is something you can see here. The
10 Pelc study required at least five days of clear
11 abstinence before they could enter the study. In
12 the German study, this was between two weeks and
13 four weeks, the window in which they could enter
14 the study. In the French study, this was one week
15 up to about four weeks, also.
16 [Slide.]
17 Here are the methods for collecting data,
18 drinking data, in these three studies in Europe.
19 Of course, there is self-report on alcohol
20 consumption at each visit done by the patient.
21 Then, of course, also, there is confirmation
22 looking at biological markers such as gamma GT,
23 liver-function test, MCV, CDT and also
24 breathylizing at each single visit. So each most
25 of these things were done at each of these single
43
1 visits.
2 Then, of course, the investigator who
3 either was a trained psychologist working in this
4 field or who was a doctor working in this field,
5 they had to make and give their clinical global
6 impression about the drinking status of the
7 patient.
8 These were in addition to their
9 professional training. They were also trained
10 prior to the studies in collecting the data using
11 the interviews and the material that was provided
12 in these studies.
13 Then, finally, family members or other
14 caretakers such as the private doctor or the family
15 doctor of the patient was also involved in trying
16 to find out what the drinking status of this
17 patient was. This was done, the integration of all
18 this material, or all this information, by the
19 investigator who had then to say, well, he is
20 abstinent or he is drinking and he did resume
21 drinking ten days ago, for instance.
22 Whenever there was a discrepancy between
23 those variables, then we said, okay, we are going
24 the conservative way. Then we say he was drinking.
25 [Slide.]
44
1 Here is the number of patients. You can
2 see we had about 1,000 who were randomized. This
3 is the ITT population. Then the completion of the
4 study you can see here. That is, to me at least, a
5 very important point. You can see that in the
6 acamprosate arms, we retained much more, or many
7 more, patients than we did in the placebo arm.
8 Also this is not, and was not, an outcome
9 criteria in the first place. To me, as a
10 clinician, as a psychiatrist, this is a very
11 important issue because as long as I have and see
12 the patients, I can do something about them. So,
13 for me, clinically, this is a very meaningful and
14 positive figure here.
15 So then, conversely, of course, we have
16 these figures of the patients who discontinued the
17 study.
18 [Slide.]
19 The reasons for discontinuation were
20 different. We had, as you have seen already, 46
21 percent who discontinued while being on acamprosate
22 and 60 percent being on placebo, for instance,
23 because of lost-to-follow-up or treatment failure
24 or other reasons such as patient refusal, et
25 cetera. So, there again, we meet these figures
45
1 that I have shown to you before. More patients
2 stay in treatment when they are treated with
3 acamprosate.
4 [Slide.]
5 Here are the demographics of these
6 studies. Of the patients in these studies, we had
7 85 or 80 percent males, more or less. We had a
8 mean age of 42, 43 years, 70 kilograms of body
9 weight. We had a mean duration of alcohol
10 dependence of about ten years throughout these
11 studies.
12 Then these are the consumption data from
13 which you can see that many of those people, like
14 around 80 percent or 70 to 80 percent, had had,
15 like, ten shots of whiskey a day, which is a lot,
16 or 40 ounces of wine a day or 80 ounces of beer a
17 day.
18 All of them had been detoxified prior to
19 the entry of the study. That is, again, something
20 we have already touched upon. So almost all were
21 abstinent at baseline. So that was the same across
22 all three European studies which I am presenting to
23 you here.
24 [Slide.]
25 Here are the treatment exposures in weeks
46
1 within these studies. You can see the Pelc study
2 which lasted 13 weeks, the average was 10 to 11
3 weeks on treatment. The German study--by the way,
4 this acronym stands for Prevention of Relapses in
5 Alcoholics with Acamprosate. That is what PRAMA
6 stands for--48 weeks, and we had about 32 weeks on
7 treatment in the acamprosate group and less on
8 treatment in the placebo group, and then, again,
9 the Paille study, 35 in the low-dose acamprosate,
10 37 in the high-dose and 31 in the placebo group.
11 [Slide.]
12 Compliance. This was based on pill count,
13 the pills that were turned in by the patients when
14 they came to the visits. So we have about 97
15 percent in the Pelc study. Because these are much
16 longer, of course, we have a lower but still
17 satisfactory compliance of 81 percent in the German
18 study, 82 to 88 percent in the French study.
19 [Slide.]
20 Here are the outcome criteria that were
21 used throughout these three European studies.
22 First of all, of course, was time to first drink.
23 So whenever someone had a relapse, we counted, or
24 we measured the time when this occurred and this
25 was entered into this analysis. Then we did a
47
1 Kaplan-Meier statistics on this.
2 Then, the second outcome criterion was
3 rate of rate of complete abstinence which meant the
4 percent of patients completing the study without
5 consuming any alcohol. Of course, these two are
6 very conservative measures and you certainly--or,
7 let's put it another way. There is more
8 information in this data than just the time when
9 someone had his first relapse because these are
10 relapses to drinking at all. That is different
11 from the studies which were done at the same time
12 in the U.S. where you had return to heavy drinking.
13 This is return to the first drink.
14 If you do this, you might lose someone who
15 had one drink or maybe two days of drinking and
16 then he was abstinent again and he is always
17 counted as a failure.
18 So, what we did in order to try to pick up
19 this additional information is we looked at
20 something that was called cumulative abstinence
21 duration in percent. That is the time on the study
22 where a patient is reported to be abstinent no
23 matter whether they had a relapse or not at some
24 time during the study.
25 [Slide.]
48
1 So here are the results. First, the
2 Kaplan-Meier for the first drink in the Pelc study.
3 You see the placebo group in blue. They are having
4 relapses. Of course, the other patients have
5 relapses. The difference between the two groups is
6 statistically significant if you take dropouts as a
7 failure.
8 There is no difference between the two
9 dosages. The low and high dosage of acamprosate
10 did not produce a significant difference between
11 those two but the other one compared with placebo
12 was clearly significant.
13 [Slide.]
14 The same is true for the PRAMA in Germany,
15 again, time to first relapse. Those on placebo,
16 they tended to relapse earlier than the patients on
17 acamprosate.
18 [Slide.]
19 The Paille study; again, we have a
20 difference between placebo and the two treatment
21 arms with acamprosate which, again, between the two
22 arms, there was not a difference in the Paille
23 study time to first relapse.
24 [Slide.]
25 The second outcome criterion, you
49
1 remember, was complete abstinence or rate of
2 complete abstinence. Here is the Pelc study,
3 again, after only three months of treatment. Those
4 on placebo had about 14 percent whereas the others
5 who were treated with acamprosate were at about 40
6 percent abstinent after twelve weeks.
7 After one year in the German study, we
8 have here 12 percent versus 29 percent, again a
9 very clear-cut 2.4-fold advantage for acamprosate.
10 In the Paille study, there is a significant
11 difference between placebo and the high dosage of
12 acamprosate, also. So, I think, also they are
13 clear results.
14 [Slide.]
15 Now this percentage of abstinent days, or
16 the CAD percent. Again, in the Pelc study we have
17 a difference between placebo and the two treatment
18 arms. In the German study, we have the same. In
19 the Paille study, placebo also is different from
20 the high-dosage of acamprosate.
21 [Slide.]
22 Here is the summary. First outcome
23 criterion, time to first relapse, a clear
24 indication that acamprosate works better with about
25 a factor of two to three times longer stay with
50
1 relapse than in the placebo group.
2 The complete abstinence rate, it is the
3 same result, between 1.7 and 2.7 times greater or
4 better with acamprosate compared with placebo.
5 Also, for this third outcome criteria, we have an
6 advantage in favor of acamprosate versus placebo.
7 These were the results of these three pivotal
8 studies which were done in Europe.
9 [Slide.]
10 With my final slide, I would like to show
11 you again where I work and I might see you again at
12 one of these occasions. Thank you.
13 Analysis of the U.S. Study Results
14 DR. MASON: Good morning.
15 [Slide.]
16 I am Dr. Barbara Mason and I served as
17 overall principle investigator for the U.S.
18 acamprosate trial.
19 [Slide.]
20 In this section, I am going to first be
21 covering these points for the U.S. multicenter
22 trial. I will conclude by integrating the U.S. and
23 European acamprosate clinical trial experience in
24 outpatients with alcohol dependence.
25 [Slide.]
51
1 The U.S. multicenter trial had two
2 overarching and somewhat competing objectives. The
3 first objective was to provide FDA with the
4 requested reassurance about the safety of
5 acamprosate in the typical American outpatient with
6 alcohol dependence who was considered to be more
7 likely to abuse other drugs and have less access to
8 inpatient detoxification services than their
9 European counterparts.
10 Additionally, because acamprosate is not
11 metabolized and it is eliminated unchanged by the
12 kidneys, there was interest in examining the safety
13 of acamprosate without any restrictions on study
14 admission because of serum-creatinine level or
15 liver-function-test abnormalities or patient age as
16 opposed to the European studies.
17 One implication, of course, of no upper
18 age limit is greater chronicity in a progressive
19 disorder and, likewise, no upper limit for
20 liver-function test may admit patients with more
21 severe alcohol dependence.
22 [Slide.]
23 The second objective related to the
24 sponsor's interest in evaluating the efficacy of
25 the standard therapeutic 2 grams per day dose of
52
1 acamprosate but given as two 500 milligram tablets
2 twice a day in contrast to the European dosage
3 schedule of two 333 milligram tablets three times a
4 day.
5 These two dosing schedules had previously
6 been shown to be bioequivalent in the multidose
7 crossover pharmacokinetic study. In addition,
8 given the safety and tolerability of the standard 2
9 gram dose of acamprosate, there was interest in
10 evaluating a higher 3 gram daily dose on an
11 exploratory basis in a smaller group of subjects.
12 [Slide.]
13 We developed two strategies specified in
14 the study protocol and case-report form to control
15 for factors generally associated with reduced
16 alcoholism treatment efficacy. The study was
17 particularly vulnerable to the influence of these
18 factors because of the broad admission criteria
19 which had been requested by the FDA for their
20 safety evaluation.
21 First, as in many pharmacologic studies
22 involving drugs with prolonged time to steady
23 state, an efficacy evaluable population was defined
24 that included those subjects who took medication
25 for the seven days needed to reach acamprosate
53
1 steady state and who were at least 75 percent
2 compliant with medication thereafter.
3 Additionally, this efficacy evaluable
4 population excluded those whose urine tested
5 positive for elicit drugs at any study visit. A
6 second strategy was to include standardized
7 baseline measure of variables identified in the
8 alcoholism-treatment literature as reliably
9 associated with poor outcome such as severity of
10 dependence or comorbidity or treatment goal of
11 nonabstinence.
12 These variables were to be examined in
13 relation to outcome as potential covariates in
14 order to reduce residual variation in the analyses
15 and to off set the influence of random imbalances
16 of baseline variables, particularly for subgroups
17 of interest.
18 [Slide.]
19 As in the European pivotal trials, the
20 U.S. study was double-blind, placebo-controlled
21 with random assignment to treatment and all
22 subjects met DSM criteria for alcohol dependence.
23 Unlike the European pivotal trials, the U.S. study
24 did not exclude substance abusers or those over
25 65 years of age and did not require detoxification
54
1 nor an abstinent interval prior to randomization.
2 In the European pivotal trials, as Dr.
3 Mann mentioned, all subjects received whatever
4 supportive psychosocial therapy was routinely used
5 by the center or investigator. Conversely, in the
6 U.S. trial, a standardized behavioral therapy
7 program that included a scripted therapist manual
8 and patient handout materials was provided to all
9 study participants.
10 [Slide.]
11 There is no gold standard for determining
12 drinking occurring between study visits or office
13 visits. Therefore, self-report with multiple
14 sources of corroboration whenever possible is the
15 current state of the art, both for alcoholism
16 pharmacotherapy trials as well as in treatment
17 settings.
18 European pivotal and U.S. trials all
19 relied on self-reported drinking gathered under
20 specific conditions shown to enhance accuracy of
21 self-report including eliciting the drinking data
22 by an alcoholism expert and providing written
23 assurance of confidentiality of the data.
24 All data were collected in clinical or
25 research settings which encouraged honest reporting
55
1 as opposed to probation offices or other settings
2 which might have legal or punitive ramifications
3 for disclosure of drinking.
4 Three of the four trials provided diaries
5 that were collected at each study visit either to
6 aid recall or to provide information on general
7 clinical status. Only the U.S. study included a
8 daily drinking calendar using standard drink icons
9 to enhance precision of self-reported quantity and
10 frequency of drinking, as shown in the next slide.
11 [Slide.]
12 Standard drinks were defined on the basis
13 of alcohol content with a beer equal to a glass of
14 wine equal to a shot of hard liquor. Although
15 standard drinks in the U.S. study contained
16 approximately 15 grams of pure alcohol, a bit more
17 generous than shown here, I am showing you these
18 12-gram icons because for today's presentation and
19 for your briefing document, all drinking
20 information is based on the smaller European
21 12-gram standard drink.
22 [Slide.]
23 All pivotal trials included multiple
24 biochemical measures to confirm validity of
25 self-report of abstinence or drinking. All trials
56
1 used gamma GT. Both the PRAMA and U.S. studies
2 breathylized patients at each study visit, and Pelc
3 II and U.S. trials tested for alcohol in urine as
4 well.
5 Additionally, PRAMA, Paille and the U.S. trials
6 verified patient self-report with a close friend or
7 relative specified by the patient at multiple time
8 points.
9 In all trials, if there were discrepancies
10 between patient self-report and the corroborating
11 information, typically the most negative outcome
12 would be assumed accurate. The drinking intervals
13 assessed in each trial were of sufficient duration
14 to capture infrequent drinkers and were consistent
15 with methodologic studies confirming the validity
16 of self-report for intervals of these durations.
17 [Slide.]
18 The primary study outcomes in the European
19 pivotal trials were informed by an
20 abstinence-oriented treatment tradition with all
21 patients undergoing detoxification and beginning
22 study participation in an abstinence state.
23 Therefore, the first information obtained from
24 participants in these trials at each visit was did
25 they or did they not drink since their last study
57
1 visit.
2 Four patients who did report drinking, an
3 effort was made in all three European pivotal
4 studies to categorize the amount of alcohol
5 consumed and the number of drinking days since
6 their last study visit as per a case-report form.
7 However, all study primary outcomes, time to first
8 drink, complete abstinence rate, point prevalence
9 of abstinence, were related to abstinence or
10 nonabstinence.
11 Consistent with clinical practice and
12 research involving alcoholism, patients who
13 discontinued prematurely due to alcohol-related
14 reasons, or patients for whom follow-up information
15 was not available were considered treatment
16 failures and as nonabstinent for the remaining
17 treatment period.
18 [Slide.]
19 Conversely, the time-line follow-back
20 method used for data collection in the U.S. trial
21 was a research tool originally developed to assess
22 continuous variables associated with controlled
23 drinking as a study outcome as opposed to the
24 categorical outcome of abstinence/nonabstinence.
25 It involves a more rigorous emphasis on
58
1 retrospective estimates of daily drinking through
2 the use of calendar-based memory aids and standard
3 drink icons to enhance recall.
4 The tradeoff for the increased precision
5 of the time-line follow-back method is that it
6 requires more time to administer thereby increasing
7 the burden on the subject in clinic personnel.
8 This may result in increased attrition rates and
9 may be inappropriate in a clinical setting where
10 time is at a premium unless more precision on
11 drinking behavior is needed.
12 In U.S. clinical practice, the time-line
13 follow-back is not used for these reasons. U.S.
14 clinical practice more directly reflects the
15 drinking data collection methods of the European
16 studies.
17 Additionally, the time-line follow-back
18 method used in conjunction with the daily drinking
19 diary, as in the U.S. study, may, in itself, reduce
20 drinking. This impact on outcome has been shown
21 for self-monitoring techniques and other
22 indications; for example, Weight Watchers.
23 One can note that, in a double-blind,
24 placebo-controlled trial, the impact of study
25 procedures should be equally distributed across
59
1 treatment groups. Nevertheless, if study
2 procedures have a therapeutic influence, then the
3 study is actually comparing background treatment
4 plus placebo to background treatment plus
5 acamprosate and the presence of the background
6 treatment might reduce the effect size for
7 potential improvement that acamprosate could
8 provide.
9 [Slide.]
10 The U.S. study was a three-armed trial
11 with subjects randomized in a 3 to 3 to 1 ratio to
12 placebo, acamprosate 2 grams a day or acamprosate 3
13 grams a day. 741 patients were screened and, of
14 these, 601 outpatients with alcohol dependence
15 representing 81 percent of those screened were
16 randomized to 6 months of treatment.
17 After the treatment phase, patients were
18 followed for an additional two months
19 off-treatment. In my discussion of the U.S. study,
20 I am going to focus on the comparison between
21 acamprosate 2 grams and placebo since that
22 comparison forms the basis of the sponsor's NDA.
23 The 3-gram group was an exploratory dose group of
24 smaller size, as you can see and I won't address it
25 further this morning.
60
1 [Slide.]
2 As I mentioned before, in comparing the
3 methodologies of the U.S. and European studies, in
4 the U.S. study, all patients were provided with a
5 brief standardized behavioral-therapy program at
6 every study visit. The program was based on
7 principles of motivation enhancement with the goals
8 of abstinence and methodology compliance and was
9 delivered by experienced nurses or counselors with
10 a bachelor's degree or higher.
11 Patients were provided with informational
12 handouts about alcohol and acamprosate. There were
13 also tips for quitting drinking and ongoing
14 self-assessment and interactive exercises
15 pertaining to their drinking behavior such as the
16 treatment goals work sheet and the treatment
17 progress summary.
18 The components of this program are
19 currently used in conjunction with acamprosate in
20 Europe--I have some of the materials here and am
21 happy to share them--and will shortly be available
22 on line at Acoweb, the Lipha website. In the U.S.
23 trial, the behavioral therapy was implemented
24 across psychiatry, alcoholism-specialty and
25 internal-medicine settings.
61
1 [Slide.]
2 The 21 participating treatment centers
3 were located throughout the United States as shown
4 in this map.
5 [Slide.]
6 As in the European pivotal trials,
7 patients were in their mid-40s at their time of
8 study entry although, in the U.S. trial, the age
9 ranged from 22 to 72 years with about 10 percent of
10 patients in their 60s and early 70s. Compared to
11 the three pivotal trials in Europe, there was
12 somewhat greater representation of females in the
13 U.S. trial. Racial distribution was roughly
14 equivalent to U.S. population norms.
15 The 2 gram acamprosate group included more
16 individuals living alone with fewer subjects
17 employed full-time and more individuals with a
18 significant psychiatric history than the placebo
19 group.
20 [Slide.]
21 Just to orient you, the clinical global
22 impression was a summary by the investigator of the
23 patient's current alcohol dependence severity with
24 7 being most severe. A score of 22 or greater on
25 the alcohol dependence scale indicates subjects
62
1 with substantial to severe lifetime alcohol
2 dependence severity.
3 For the measures shown here and the
4 measures of psychosocial support shown on the
5 previous slide, although generally comparable
6 across the two treatment groups, you might notice
7 that, for each variable, the 2 gram group has
8 evidence of slightly greater severity of alcohol
9 dependence than the placebo group.
10 [Slide.]
11 There was a higher proportion of patients
12 in the placebo group having a baseline goal of
13 total abstinence and a higher proportion in the 2
14 gram group requiring medicated detoxification prior
15 to study entry. Accordingly, in the aggregate,
16 subjects assigned to the 2 gram group appear to
17 have entered the trial relatively disadvantaged.
18 [Slide.]
19 As you can see, approximately
20 three-quarters of the sample reported lifetime
21 experience with illicit substances with
22 approximately one-third reporting illicit substance
23 abuse in the year prior to randomization.
24 [Slide.]
25 Slightly less than half of the population
63
1 were current smokers and between 6 and 8 percent
2 had positive urine for cannaboids at screening.
3 [Slide.]
4 You have seen the formal patient
5 disposition in your briefing document. I would
6 like to highlight certain features of patient
7 participation that may be relevant for
8 understanding efficacy. You will note high rates
9 of methodology compliance across all treatment
10 groups. However, the 2-gram group had fewer weeks
11 on study and a lower rate of study completion than
12 the placebo group. In an effort to understand this
13 further, a blinded panel of experts evaluated all
14 premature terminations in terms of alcohol
15 relatedness taking into account all available
16 information.
17 Of those patients terminating early, the
18 reason was more likely to be alcohol-related in the
19 placebo group than in contrast to the 2 gram
20 acamprosate group. There was no difference in the
21 percentage of patients across the groups for
22 terminations due the adverse events.
23 [Slide.]
24 As the FDA pointed out in their
25 information package, in contrast to the European
64
1 studies, half the U.S. study population was still
2 drinking at randomization. Therefore, the plan for
3 European-based variables such as time to relapse
4 and rate of complete abstinence became relatively
5 meaningless.
6 Similarly, as pointed out earlier, the
7 fact that the 2 gram group had briefer time on
8 study would negatively impact on their cumulative
9 abstinence duration with missing time accounted for
10 as drinking time. Furthermore, the unfavorable
11 baseline imbalances for the 2 gram group were also
12 found to meaningfully influence study outcomes.
13 [Slide.]
14 The variables that we chose to measure in
15 a standardized manner at baseline in the
16 case-report form included a brief screen of major
17 psychopathology as greater psychiatric severity has
18 been reliably associated in the literature with
19 poor alcoholism treatment outcome.
20 Although subjects with current dependence
21 in illicit substances were excluded from study
22 admission, subjects with substance abuse including
23 those with urines positive for cannabis at
24 screening at baseline were admitted to the study.
25 Given the well-known association of drug abuse with
65
1 premature treatment termination and poor alcoholism
2 treatment outcome, the illicit drug use index
3 developed by the National Institute on Drug Abuse
4 was used to characterize severity of substance
5 abuse.
6 Additionally, the Fagerstrom test of
7 nicotine dependence was used to capture current
8 severity of nicotine dependence.
9 As with psychiatric and substance-abuse
10 comorbidity, greater severity of alcohol dependence
11 has generally been associated with poor treatment
12 response especially for outpatients. In the
13 American study, current severity of alcohol
14 dependence was assessed with the investigator's
15 clinical global impression and lifetime severity
16 with the Alcohol Dependence Scale.
17 [Slide.]
18 Fewer social supports result in worse
19 treatment response generally but especially in the
20 case of outpatient treatment of alcoholism.
21 Readiness to change emerged as the strongest
22 predictor of long-term drinking outcome in Project
23 MATCH and, as in Project MATCH, was measured, in
24 this study with DiClementi's stages of readiness to
25 change.
66
1 Initial commitment to complete abstinence
2 has been shown to predict higher rates of
3 abstinence among alcoholics, opiate users and
4 cigarette smokers. In contrast, subjects having a
5 goal of minimizing a slip or having other drinking
6 goals are typically more likely to relapse.
7 Treatment goals were assessed at baseline
8 in the U.S. study with a standardized treatment
9 goals check list which I will be showing you.
10 Compliance with prescribed treatment has been
11 significantly associated with drinking outcome in
12 both behavioral and pharmacological clinical
13 trials.
14 In the U.S. study, methodology compliance
15 was estimated on the basis of pill count from
16 returned blister packs at every study visit.
17 Ingestion of acamprosate was verified by plasma
18 acamprosate levels at week 1 and end of study
19 although results were not available until after
20 study unblinding.
21 Importantly, and finally, as Babour and
22 colleagues have pointed out, these factors may be
23 most meaningfully used in combination to create a
24 multidimensional model to understand alcoholism
25 treatment outcome.
67
1 [Slide.]
2 The FDA has requested that we provide an
3 analysis to reconcile the findings of the U.S. and
4 European trials and to further our understanding of
5 how acamprosate would be beneficial in American
6 alcoholics. Given that missing data are attributed
7 to relapse in study-outcome calculations, in order
8 to better understand the efficacy of the 2 gram
9 group, a standard panel of covariates relating to
10 baseline measures of psychosocial support and
11 alcoholism severity and treatment exposure were
12 uniformly applied to all outcome measures.
13 Statistical modeling associated early
14 termination with baseline variables relating to
15 psychosocial support and disease severity rather
16 than to treatment group assignment.
17 [Slide.]
18 The actual chest list used in the case
19 report form to capture patients treatment goals at
20 baseline is depicted in this slide. In the FDA's
21 analysis, patients with the goal of abstinence were
22 grouped together with those acknowledging that they
23 could have a slip and the difference in results may
24 serve to emphasize the importance of complete
25 commitment to abstinence at treatment onset to
68
1 optimize acamprosate efficacy.
2 [Slide.]
3 Because the first dose of study
4 methodology was an observed dose given in a clinic,
5 all 601 randomized patients were included in the
6 safety population. The intention to treat, or ITT,
7 population represented all randomized patients for
8 whom any follow-up efficacy data were available. I
9 have already described to you the a priori defined
10 efficacy evaluable population.
11 As Sharon Hall and colleagues at the
12 University of California and Stephanie O'Malley and
13 colleagues at Yale University have reported,
14 commitment to total abstinence is related to a
15 lower risk of returning to use of alcohol as
16 opposed to goals that include slips, controlled
17 drinking or other levels of alcohol consumption.
18 One of the DSM-IV diagnostic criteria for
19 alcohol dependence specifically relates to the
20 tendency to drink more than originally intended.
21 Consequently, complete abstinence is the treatment
22 goal recommended by NIAAA and other expert groups.
23 Because a treatment goal of abstinence was
24 so strongly associated with positive U.S. study
25 outcomes, subjects within the ITT and efficacy
69
1 evaluable population who, at baseline, identified
2 their treatment goal as total abstinence were
3 looked at as additional subpopulations in order to
4 better understand acamprosate efficacy in the U.S.
5 population.
6 We have called these subpopulations
7 respectively the motivated ITT population and the
8 motivated efficacy evaluable population.
9 [Slide.]
10 Cumulative abstinence duration or percent
11 of abstinence time on study was the only original
12 outcome parameter which was still applicable to the
13 U.S. study population since it does not involved
14 censoring of data at the time of the first drink.
15 In the original European-based analysis plan for
16 the calculation of this outcome parameter, the
17 number of abstinent days were divided by the total
18 duration of the trial.
19 Given the precision of the U.S. data
20 collection and follow-up methods in the revised
21 analysis the denominator remained the total trial
22 duration unless patients were censored for leaving
23 the trial for reasons unrelated to alcohol.
24 Also, as stated earlier, in order to
25 better understand the efficacy of acamprosate in
70
1 the U.S. population, a standard panel of baseline
2 and treatment exposure covariates would uniformly
3 applied across outcome measures in order to reduce
4 residual variation and offset the imbalances in
5 comparisons between acamprosate and placebo.
6 This adjustment enables the supportive
7 identification of trends with p less than 0.05 in
8 favor of acamprosate 2 grams relative to placebo in
9 the ITT group. The extent of these favorable
10 trends increases as one moves to the more defined
11 populations mainly because of the larger increase
12 in cumulative abstinence duration percent in the
13 acamprosate group than in the placebo group.
14 Abstinence time was about 6 percent longer
15 with acamprosate 2 grams in the ITT population
16 while for patients in the efficacy evaluable
17 population who had total abstinence as their
18 treatment goal, abstinence time was about 16
19 percent longer with acamprosate 2 grams relative to
20 placebo.
21 This supports the premise that motivation
22 to be abstinent merits consideration in the
23 interpretation of acamprosate efficacy in the U.S.
24 population.
25 [Slide.]
71
1 Furthermore, the previously noted trends
2 with acamprosate were maintained in the 2 gram
3 group relative to placebo during the two months
4 post-treatment follow-up phase again most markedly
5 in those subjects with a baseline motivation of
6 total abstinence.
7 [Slide.]
8 As support analyses of cumulative
9 abstinence duration, covariate adjusted odds ratios
10 were calculated for the likelihood of good response
11 with acamprosate relative to placebo. Good
12 responders were defined as those subjects with a
13 cumulative abstinence duration of 90 percent or
14 more. This is highly relevant from a clinical
15 point of view.
16 For the motivated efficacy evaluable
17 population, the adjusted odds ratio for good
18 response with acamprosate versus placebo
19 supportively had p less than 0.05 and corresponded
20 to about three times higher odds for good response
21 with acamprosate 2 grams than with placebo.
22 Conversely, poor response was defined as those
23 subjects having a cumulative abstinence duration of
24 10 percent or less.
25 The adjusted odds ratio for poor response
72
1 for acamprosate 2 grams relative to placebo had p
2 less than 0.05 and showed a decreasing pattern of
3 lower odds for poor response for acamprosate 2
4 grams across the subgroups.
5 [Slide.]
6 A final support analysis of abstinence
7 looked at the likelihood of a subject being
8 abstinent during the interval prior to their last
9 treatment-phase visit. This outcome may have
10 clinical relevance in that a subject's behavior at
11 the end of study may be predictive of behavior off
12 study.
13 There was a trend for subjects treated
14 with acamprosate 2 grams to have a high odds for
15 being abstinent at the end of study participation
16 compared to placebo with compliant and motivated
17 patients having more than twice the odds to be
18 abstinent at this key time point in the 2 gram
19 group.
20 [Slide.]
21 Now I am going to turn to a secondary
22 outcome that involves quantity of drinking on
23 study. The calendar method of drinking data
24 collection in the U.S. study permitted the most
25 detailed examination to date of whether acamprosate
73
1 reduces alcohol consumption in nonabstinent
2 subjects during the study.
3 You will recall that all subjects received
4 a standardized alcohol-specific behavioral therapy
5 and, as this slide shows, all patients, including
6 the placebo group, showed substantial reductions on
7 study from baseline levels of drinking. However,
8 particularly in those subjects motivated to be
9 abstinent, the covariate adjusted analysis showed a
10 larger reduction with acamprosate 2 grams than with
11 placebo.
12 Again moving from ITT to the more defined
13 subpopulations, there was a further reduction of
14 only approximately 3 percent in the placebo group
15 compared to almost 20 percent in the acamprosate 2
16 gram group. This provides further support for an
17 association between motivation to be abstinent and
18 trends in favor of acamprosate relative to placebo
19 in the U.S. population.
20 [Slide.]
21 As seen in this slide, all treatment
22 groups showed an improvement in mean levels of GGT
23 at study endpoint relative to the elevations in
24 mean values seen at baseline. Mean endpoint values
25 were normal or near normal in this predominantly
74
1 male study population further attesting to the
2 improved status of patients in all groups and the
3 validity of self-report in this study.
4 [Slide.]
5 As requested by the FDA, the U.S. study
6 population was much more inclusive than seen in
7 most clinical trials in alcohol dependence in order
8 to assess the safety of acamprosate in patients
9 with polysubstance abuse, hepatic and renal
10 dysfunction and the elderly.
11 As a result of the U.S. study's broad
12 admission criteria, 81 percent of screened patients
13 were randomized supporting the external validity of
14 the study. I want to emphasize that, in contrast,
15 in an ongoing large multicenter trial in
16 alcohol-dependent patients, only about 25 to
17 30 percent of screened patients were randomized.
18 The rate of compliance with medication
19 exceeded 88 percent in all treatment groups lending
20 support to the acceptability of both acamprosate
21 and the divided dosing schedule.
22 [Slide.]
23 Controlling for baseline variables in
24 treatment exposure, the U.S. study results support
25 the efficacy of acamprosate 2 grams relative to
75
1 placebo particularly in patients with a baseline
2 goal of abstinence. This treatment group had
3 increased cumulative abstinence duration and
4 increased likelihood of good response, a decreased
5 likelihood of poor response and an increased
6 likelihood of being abstinent at study termination.
7 In addition, although all groups showed
8 improvement in drinking behavior on study relative
9 to baseline, the 2 gram group had a greater
10 decrease in both the quantity and frequency of
11 alcohol consumption compared to placebo.
12 Self-reported drinking were confirmed by
13 accompanying changes in GGT. A consistent
14 association was found between trends in favor of
15 acamprosate and a baseline goal of total abstinence
16 across study outcomes.
17 This observation has implications for
18 healthcare providers prescribing acamprosate for
19 their outpatients with alcohol dependence.
20 [Slide.]
21 Integrating the U.S. and European
22 pivotal-trial exposure with acamprosate, overall,
23 acamprosate 2 grams per day showed significant
24 effects on abstinence outcomes in almost 2000
25 alcohol-dependent outpatients participating in
76
1 double-blind, placebo-controlled trials up to one
2 year in duration.
3 Additionally, acamprosate showed continued
4 efficacy during off-treatment follow-up periods of
5 as long as one year.
6 [Slide.]
7 European and U.S. data suggest that
8 acamprosate does not induce abstinence in
9 unmotivated drinkers. In Europe, patients had to
10 make a commitment to abstinence-oriented treatment
11 that began with formal detoxification typically
12 inpatient. Thus, their treatment goal at the onset
13 of the clinical trial was implicitly total
14 abstinence and treatment effects may have been
15 easier to discern because of the resultant
16 homogeneity.
17 In contrast, the U.S. study population did
18 not typically undergo detoxification and was quite
19 heterogeneous in their expressed baseline treatment
20 goals. Through examination of subpopulations,
21 defined by the presence of total abstinence as a
22 treatment goal, the U.S. data suggest that it is
23 not necessary to undergo formal detoxification in
24 order to obtain therapeutic benefit from
25 acamprosate provided patients are motivated for
77
1 total abstinence.
2 [Slide.]
3 Uniformly, high rates of compliance across
4 the pivotal trials support the acceptability of
5 acamprosate in the twice daily and three times
6 daily dosing schedules used in these studies. The
7 pivotal trials spanned a range of countries and
8 clinical settings. You may also recall that the
9 European studies, by design, did not include any
10 uniform behavioral therapy. Thus, the efficacy of
11 acamprosate is supported across a broad range of
12 treatment orientations.
13 Closing Remarks
14 DR. GOODMAN: Ladies and gentlemen,
15 members of the committee, I would like to spend
16 these final few minutes of our presentation on the
17 issues set forth by the division for your
18 consideration.
19 [Slide.]
20 In our briefing document and in our
21 presentations this morning, we have described to
22 you three European double-blind, placebo-controlled
23 studies of acamprosate that meet all the FDA
24 criteria for approvability. As we are all aware,
25 the process of drug development is a long one, more
78
1 often than not. The European studies on which we
2 are relying as evidence of efficacy were conducted
3 starting in 1989 and were completed for the most
4 part by 1995.
5 Although the FDA has characterized these
6 studies as older, in fact, the trials were
7 conducted by qualified clinical experts in the
8 field of alcoholism. They meet the FDA criteria of
9 being clinically generalizable to the target
10 population in the United States and the trials were
11 conducted in a manner consistent with good clinical
12 practice and are auditable.
13 [Slide.]
14 Despite differences of opinion about the
15 most appropriate methodology for assessing outcome
16 in clinical trials of alcohol dependence, and, in
17 fact, despite the actual methodologies applied,
18 those of Lipha or those of the FDA, the three
19 European pivotal trials showed consistently
20 significant and clinically relevant effects both on
21 parameters selected as primary, as shown here, as
22 well as various secondary parameters described in
23 your briefing document.
24 These studies, along with the others
25 described in the documents provided to you, served
79
1 and continue to serve as the basis of regulatory
2 approvals around the world, most recently in
3 Australia and South Africa. The results of the
4 European trials are applicable to approvability for
5 the United States because there is no biologic or
6 pharmacokinetic reason to believe that drug
7 response in alcoholic patients will differ between
8 Europe and the United States.
9 As Dr. Koob has pointed out, the drug is
10 not metabolized. Nor is there any reason to
11 believe that the nature of alcohol dependence
12 differs in European and American alcoholic
13 patients.
14 [Slide.]
15 In a letter to the FDA, at their request,
16 from the National Institute on Alcohol Abuse and
17 Alcoholism specifically addressing this issue, the
18 concluding comments from NIAAA are the core illness
19 of alcohol dependence is similar in the United
20 States and Europe.
21 The conclusion is based on several
22 considerations. First, the diagnostic methods for
23 coding alcohol dependence are very similar in the
24 United States and Europe. Most of the clinical
25 trials of acamprosate in Europe used the Diagnostic
80
1 and Statistical Manual of Mental Disorders III-R,
2 the DSM III-R, for verification of alcohol
3 dependence, the version which was available and
4 used in both Europe and the United States at the
5 time these studies were conducted.
6 Second, an international conference, held
7 in Germany in September of 1999 to determine if
8 cross-national studies could be conducted,
9 concluded that, while there are cultural
10 differences between the U.S. and Germany,
11 cross-national collaboration was feasible because
12 the alcohol-dependent populations were similar.
13 Finally, a comparison of the cardinal
14 symptoms of the alcohol dependence syndrome in U.S.
15 and Soviet populations revealed virtually identical
16 characteristics.
17 [Slide.]
18 Dr. Barbara Mason has shown you, through
19 analyses using an informed set of baseline
20 variables and treatment exposure that the American
21 study results are not in conflict with the European
22 experience when baseline differences among the
23 treatment groups are controlled for. In fact,
24 these analyses have led to a further understanding
25 of the sorts of patients who might ultimately
81
1 benefit from acamprosate, namely patients who are
2 motivated to total abstinence without a slip.
3 These data are offered to you not as a
4 justification of the methods and the results but,
5 instead, as an explanation for what happened in the
6 very broadly inclusive U.S. trial and as a means of
7 assuring you that the populations are, indeed,
8 similar.
9 Dr. Mason has presented to you the
10 interpretation of these additional analyses which
11 showed that acamprosate increased the percentage of
12 abstinent time on study, increased the likelihood
13 of remaining abstinent for 90 percent or more of
14 the time on study and, as shown in your briefing
15 document, also impacted favorably on alcohol
16 consumption in those patients who did drink.
17 [Slide.]
18 Taken as a whole, the acamprosate clinical
19 data submitted to the FDA for the indications shown
20 here which are being considered in the context of
21 an accelerated review more than meet the FDA's
22 criteria for approval and do not warrant additional
23 safety and efficacy trials.
24 It would unduly penalize that percentage
25 of alcoholic patients who may benefit from
82
1 acamprosate as well as their families and the
2 community in general if approval were to be delayed
3 while we all await further studies and analyses.
4 We acknowledge that acamprosate is not the
5 magic bullet we all seek for just about any medical
6 condition you could describe. However, the overall
7 picture is, indeed, clear enough both from the
8 perspective of acamprosate's efficacy and safety to
9 proceed further with the approval process.
10 [Slide.]
11 The FDA agreed that acamprosate deserved
12 an expedited review when our NDA was filed and
13 their ongoing review of the extensive data
14 submitted has been very thorough in this
15 therapeutic area in which they are practically
16 pioneers.
17 The paucity of available therapies for the
18 treatment of alcoholism and the continued enormity
19 of the personal and economic costs of alcoholism
20 mitigate, however, for action now rather than
21 later.
22 Thank you very much.
23 Questions from the Committee
24 DR. OREN: We now turn to the portion of
25 meeting where the members of the committee have the
83
1 opportunity to question Lipha about their
2 presentation or anything else with regard to
3 today's questions.
4 Given that there are millions of Americans
5 who may be affected by our recommendation, I
6 encourage our committee members not to be shy but
7 to be very vocal in coming up with questions.
8 Anyone wish to begin?
9 Dr. Keck?
10 DR. KECK: I know our task is mostly
11 around efficacy today but I have some questions
12 just about safety. It is likely that a lot of
13 people, say, with bipolar disorder who have high
14 rates with alcoholism could take this drug. What
15 do we know about drug interactions with lithium or
16 NSAIDs or other drugs that are renally cleared?
17 DR. GOODMAN: I can tell you that, from a
18 formal point of view, we have not conducted any
19 pharmacokinetic interaction studies with those
20 classes of drugs and there could be some reason to
21 suspect, mechanistically, that they might interact.
22 But we don't have the information to date.
23 I don't know, Dr. Chabac, if you are aware
24 of any patients in our postmarketing
25 pharmacovigilence database that might have been
84
1 exposed to lithium?
2 DR. CHABAC: As I told you, we have 1.5
3 million patient year exposure. You know well that
4 those patients probably have high comorbidity and
5 were treated with these kinds of products. On
6 postmarketing surveillance, we had no specific
7 problem, specific interaction with those drugs.
8 But, as Dr. Goodman told you, we didn't investigate
9 all possible drugs to be associated with
10 acamprosate.
11 DR. GOODMAN: I might add one thing, but I
12 don't want to go out of the boundaries of our
13 restricted discussion of efficacy. But we did look
14 at NSAIDs or analgesics in general in terms of
15 adverse-event occurrence in the U.S. study. My
16 recollection is that there was no difference in
17 pattern of adverse events or increased incidence.
18 But that is just based on--we have been focussing
19 on efficacy both for the preparation of this
20 meeting and so I would want to verify that. It's a
21 good point.
22 DR. KECK: Just one other basic
23 pharmacokinetic question. It was unclear to me in
24 how many of the studies the recommendation was that
25 the drug be taken with food. But that struck me as
85
1 curious since, if I am reading the data correctly,
2 food decreases the absorption and bioavailability
3 of a drug that already has limited bioavailability.
4 Can you help me understand that?
5 DR. GOODMAN: It's a good point. My
6 interpretation of that would be, again, that this
7 is not a drug that you are taking acutely such as
8 you might use an NSAID for a headache or joint pain
9 or whatever. It is something that is chronically
10 administered.
11 So I think the food effect really is of
12 minimal importance over the long haul once a person
13 is at steady state. Dr. Porte may have some
14 thoughts about that as well. She is our
15 pharmacokineticist from Lyon.
16 DR. PORTE: To answer your question, the
17 food interaction study was performed for a single
18 dose administration. This does not correspond to
19 the dosing schedule recommended in the labeling.
20 So we expect that this food interaction will not
21 impact on the clinical efficacy of this compound
22 even thought the bioavailability is already
23 limited.
24 DR. GOODMAN: As far as the clinical
25 trials are concerned, and Karl Mann may have some
86
1 comments on that, in many instances, the drug was
2 taken with meals as a reminder for taking--
3 DR. OREN: Dr. Winokur?
4 DR. WINOKUR: I wanted to ask a question
5 related to difference between the populations
6 included in the European trials and the U.S. trial.
7 If I remember data from the packet, which I don't
8 think was commented on, a considerably higher
9 percentage of subjects in the European trials had
10 histories of very high drinking histories, for
11 example greater than ten drinks a day. I just
12 wondered if some additional comment about that
13 aspect of different profiles--we have talked about
14 differences, for example, that there was other drug
15 use in the U.S. trial but I was interested in the
16 analysis of the history of drinking frequency in
17 the U.S. trial.
18 DR. GOODMAN: I will just comment. It
19 was, I won't say misrepresented in the briefing
20 document, but, in fact, the calculations presented
21 for the European data were based on drinks per day
22 for patients who did drink whereas the U.S. data
23 was shown as drinks per day for all patients,
24 whether they drank or not. So the data that Dr.
25 Mason presented in her demographics, in fact, was
87
1 the correct representation for the purposes of
2 comparison because they are basically they same.
3 DR. HUGHES: First of all, before I make
4 my comments, I just want to clarify that the
5 University of Vermont was the site of the U.S.
6 study but I did not participate in that.
7 What I wanted to ask was if there is a
8 subset of more motivated patients that acamprosate
9 works in, two things to judge post hocs on are
10 reproducibility and plausibility. So the two
11 questions I have are are there any instances,
12 either with alcoholism or other drug dependencies,
13 where a subset of more motivated patients changes
14 not the outcome but the odds ratio. That is the
15 first question. So is there a precedence for this.
16 The second question is what would you
17 maintain is the behavioral or biological mechanism
18 by which being more motivated would change, again,
19 not the outcome but, by being more motivated, it
20 would change the relative efficacy of acamprosate
21 to placebo. So, again, what is the reproducibility
22 of this and what is the mechanism?
23 DR. MASON: John, the two papers that I
24 pulled which Sharon Hall and colleagues at the
25 University of California and Stephanie O'Malley's
88
1 naltrexone study didn't calculated odds ratios.
2 They just are descriptive statistics. So I don't
3 know that I have information that would be helpful.
4 DR. HUGHES: My recollection of those
5 papers was that the more motivated did better but,
6 by being more motivated, it didn't change your
7 response to the treatment. Is that correct?
8 DR. KOOB: The more motivated did better
9 without changing their response to the treatment
10 DR. HUGHES: If you had some measure of
11 active to placebo in a study, I agree with you,
12 being more motivated is going to take your
13 abstinence rates up. But my worry is it going to
14 take them both up and not change the relative rates
15 of outcome because active and placebo.
16 I am trying to think of a prior study in
17 which, if you took a more motivated group, it
18 changed the differences between active and placebo.
19 I was wondering if you know of one.
20 DR. MASON: In the O'Malley study, which I
21 am more familiar with, it depends on what outcome
22 you look at because in the group that had the
23 behavioral therapy in which a slip was considered
24 likely, permissible, et cetera, they did have more
25 days on which drinking occurred as opposed to
89
1 patients in the behavioral therapy group that were
2 told, you must be completely abstinent.
3 Then that was crossed with naltrexone. So
4 you did have the influence of the instruction to be
5 abstinent or have a slip interacting with an
6 outcome parameter and drug. Does that help?
7 DR. HUGHES: Any thoughts about mechanism?
8 DR. MASON: George has a thought. Good.
9 DR. KOOB: I think the animal data
10 suggests that acamprosate--and this is part of the
11 pharmacokinetic issue as well. Acamprosate takes a
12 while to reach steady state. It is five to seven
13 days. Any mechanism, whether it is cognitive or
14 whether it is induced by the European studies where
15 a person had the detox, that lengthens the time
16 between when an individual has stopped drinking and
17 the onset of steady state blood levels of
18 acamprosate is going to facilitate the
19 normalization of the neurotransmitter systems that
20 it works on.
21 So my answer to that question would be
22 anything that lengthens the time that the organism
23 is without alcohol, and in that alcohol deprivation
24 study where we see a large effect, those animals
25 are not allowed to drink during the period that
90
1 they are getting acamprosate, you see a bigger
2 effect of acamprosate.
3 DR. OREN: Dr. Ortiz?
4 DR. ORTIZ: My question is Dr. Mann
5 mentioned that the European studies all had
6 psychosocial treatment programs. Dr. Mason just
7 briefly mentioned something about behavioral
8 treatment in the American study, and I am wondering
9 if we can get a little bit of elaboration on that.
10 DR. MANN: In the European studies, there
11 were not psychotherapies or psychosocial treatment
12 which was manual based. It was the treatment that
13 was, at that time, given at these different centers
14 and this might have different--within the study,
15 from center to center. So it was the center-based
16 treatment approach like counseling or, in some
17 centers, behavioral treatment. In others, it might
18 have been something else.
19 So there was not a manual-based treatment
20 in the European studies and that was different in
21 the U.S. study.
22 DR. MASON: In the U.S. study, there was a
23 manual that actually had a script in it for the
24 therapist to model and there was a training video.
25 It was based on principles of motivation
91
1 enhancement, particularly the manual developed by
2 the National Institute on Alcohol Abuse and
3 Alcoholism from Project MATCH and it also included
4 elements of the National Institute on Alcohol Abuse
5 and Alcoholism brochure for primary-care providers
6 in their approach to treating alcoholism.
7 It was really conceptualized as something
8 that could fit easily into a variety of treatment
9 settings. In fact, we deliberately included
10 internal-medicine sites. It was brief. It was
11 about twenty minutes. It could be delivered by a
12 nurse or an experienced counselor with a bachelor's
13 degree.
14 It involved handouts to the patients that
15 gave them information, let them do self-monitoring
16 exercises and really built on the patient's own
17 experience, what has worked for you in the past,
18 what hasn't worked. Then, if there is a report of
19 a drinking episode, what worked, what didn't work,
20 what do you see as the obstacles to your meeting
21 your treatment objectives.
22 Also, information like GGT levels were
23 shared with the patient so that they received
24 feedback about the progress that their efforts were
25 having in terms of the effect on their health. For
92
1 example, one of the motivation-enhancing
2 strategies, the time-line follow-back, quantifies
3 the amount of drinking that occurs in a week, for
4 example
5 Each standard drink roughly is equivalent
6 to 100 calories and people were drinking, on
7 average, about 40 drinks a week. So, when you do
8 that multiplication, people are kind of horrified
9 about how many calories are being consumed in
10 alcohol.
11 Also, another strategy is multiplying the
12 number of drinks per week by the cost. If you are
13 drinking in a bar and paying, like, $5.00 a drink,
14 people then get thunderstruck at how much they are
15 paying for alcohol. So those are some of the
16 motivation-enhancement characteristics of the
17 standardized therapy that are tracked in the
18 treatment progress summary at each visit which
19 occurs on a monthly basis.
20 Initially patients are seen one week after
21 starting medication, then, in two weeks, and then
22 they switch to the monthly schedule.
23 DR. OREN: Dr. Schatzberg.
24 DR. SCHATZBERG: This is a question for
25 Barbara Mason and George Koob. When you are
93
1 looking at the U.S. study versus the European
2 studies and you are going to a new formulation, and
3 the fact that if you look at the average weight in
4 the U.S. study, it is about 10 percent higher than,
5 let's say, in the German study that Dr. Mann talked
6 about.
7 Are we sure that we just haven't
8 underdosed in the U.S. and have you looked at
9 acamprosate levels to ascertain whether, in fact,
10 we have an effective dose in Europe in 1998 that
11 may not be effective in the U.S.
12 DR. GOODMAN: I am going to just intervene
13 even though I am not Dr. Mason. I do want to
14 correct one thing that might be a misperception on
15 the part of the committee members. These are not
16 different tablets. The only thing that is
17 different--there is no difference in the
18 formulation. They are identically formulated
19 except for the tablet strength. So there is no
20 difference in the tablet formulation.
21 As I believe Barbara pointed out in her
22 talk, there was a pharmacokinetic study,
23 multiple-dose pharmacokinetic crossover design, of
24 these two schedules which were shown to be
25 bioequivalent. So we feel, from the basis of that,
94
1 that the dosing is equivalent.
2 With regard to the question about the
3 heavier, huskier American population, I would say
4 that the Germans probably aren't too far off from
5 the American population. If you noticed in the
6 demographics, they weighed a bit more, on average,
7 certainly, than the French. So there could be a
8 comparability there.
9 We did do blood levels of acamprosate,
10 blinded of course, in the U.S. study one week after
11 starting treatment and then again at the time of
12 termination. Those blood levels were consistent
13 with steady-state levels in PK studies in our
14 dossier.
15 DR. OREN: Dr. O'Brien?
16 DR. O'BRIEN: My question also concerns
17 psychotherapeutic intervention. One of the
18 difficulties in interpreting efficacy studies in
19 any behavior disorder, whether it is depression,
20 anxiety or alcoholism is that the patients are
21 always getting two effective treatments; namely,
22 psychotherapy and a potentially effective
23 medication.
24 We have some evidence from other evidence
25 from other forms of substance abuse that there is a
95
1 dose-response curve for psychotherapy. In other
2 words, if you randomly assign people to various
3 levels of psychotherapy, it doesn't matter very
4 much which type of psychotherapy, so it is not
5 specific to, say, supportive-expressive versus
6 cognitive-behavioral. But the quantity is a
7 factor.
8 In clinics, there is a tendency, when a
9 patient is doing badly, you don't know whether they
10 are on drug or placebo but to enhance the amount of
11 time that is given to them, more frequent visits,
12 perhaps, or spending a little more time with them
13 or helping them a little bit more because you are
14 trying to help the patient.
15 I just wonder whether in either the
16 European, or any of the European studies or the
17 American study, whether there was an effort to
18 measure the quantity of psychotherapeutic
19 interaction.
20 DR. GOODMAN: I think I can answer and say
21 yes, there was. But, Barbara, maybe you want to
22 address it more specifically and Karl as well.
23 DR. MASON: I will just tell you that, in
24 the U.S. study, patients were allowed only two
25 emergency visits in addition to their monthly
96
1 visits. That was the protocol. If they required
2 more help than that, they were terminated as
3 treatment failures.
4 The treatment was proscribed to be twenty
5 minutes. It was very defined in the manual, the
6 procedures, and involved completing things and
7 reviewing things together. That was the parameter
8 of the therapy. It was adhered to.
9 DR. MANN: I think that is a very, very
10 important point which may, indeed, help to
11 understand the differences because, in the European
12 studies, the doses of psychosocial treatment was
13 extremely low. We gave very little, only a few
14 visits throughout the whole year; for instance, in
15 Germany, I think eight or nine visits.
16 So one per month in the first three months
17 and then only one every third month which is really
18 very little. So the doses which we applied were
19 really small. If you give much more--we have seen
20 this in other studies, at least we have the
21 impression that if you have a very high placebo
22 response because you give a lot of psychosocial,
23 then the drug has a harder time showing an effect.
24 If I may add something to the other
25 question earlier about the difference between the
97
1 treatment sites or the centers, we have looked,
2 because there were different forms of psychosocial
3 treatments, the differences in outcome between
4 these centers and we didn't find any difference
5 there.
6 Also, there was a difference maybe in the
7 approach psychosocially. It didn't affect the
8 overall treatment outcome. I forgot this earlier.
9 DR. GOODMAN: Barbara may want to address
10 something about the phase IV European studies, the
11 Need Project.
12 DR. MASON: There was a large
13 multinational open-label study of acamprosate that
14 was conducted specifically to look at acamprosate
15 efficacy across five major types of psychotherapy,
16 group therapy--Silvie, do you remember what some of
17 the other components were? This was in Europe.
18 There was no change, no significant difference in
19 acamprosate efficacy across the five major models
20 of psychosocial treatment that were studied.
21 This involved approximately 1200
22 outpatients with alcohol dependence all of whom
23 received acamprosate. The varying factor was the
24 co-occurring psychosocial therapy.
25 DR. OREN: Dr. Rudorfer?
98
1 DR. RUDORFER: If I could go back to the
2 pharmacokinetics for a second, given acamprosate
3 18-hour half-life, I wonder why some of the
4 European studies used three times a day dosing.
5 DR. PORTE: Actually the absorption
6 process of acamprosate is very slow so when you
7 measure the half-life, it corresponds to the end of
8 the plasma profile. Indeed, it does not correspond
9 in the case of acamprosate to pure elimination but
10 there is still some remaining absorption of
11 product. Therefore, to find a dosing regime for
12 acamprosate, we should more look at the elimination
13 half-life for the intravenous dose which is from
14 five to seven hours.
15 DR. OREN: Dr. Leon?
16 DR. LEON: In designing and implementing a
17 clinical trial, we typically take many safeguards
18 to minimize the bias of the treatment effect. What
19 we usually focus on are randomization and blinding
20 and statistical strategies. What I am struck by
21 here is it appears another strategy that is very
22 important in minimizing bias is that we prespecify
23 our primary dependent variable, our efficacy
24 measure, and prespecify our primary data analytic
25 technique.
99
1 I don't see any examples of that in these
2 four trials. It looks as if the primary data
3 analytic technique that was specified in the
4 protocol was not adhered to nor was the primary
5 efficacy measure.
6 If those measures and techniques are
7 prespecified, then anyone who looks at the data
8 will get the same answer. But when they are not
9 prespecified and changed after the data have been
10 collected, that objectivity or agreement across
11 independent assessors is jeopardized.
12 I wondered if you have a comment on that.
13 DR. G. COOK: Gary Cook. I am a
14 consultant to Lipha. I will probably need Dr.
15 Goodman's help on this. I believe that for the
16 European studies there was some type of reasonable
17 statistical plan and that the analyses that were
18 done to support the efficacy of those studies was
19 reasonably consistent with that plan.
20 There might be some further clarification
21 as to exactly what the plans were, but my
22 understanding is that there was a plan, that the
23 results were consistent with that plan and then a
24 variety of additional analyses have been done to
25 further support the robustness of the analyses of
100
1 those studies.
2 Now, the U.S. study, the issues are
3 totally different. But we need to sort of deal
4 with this in two steps so could you first clarify
5 what would have been the response to this question
6 with respect to the European studies.
7 DR. GOODMAN: I would make two points, and
8 there may be additional members of our group
9 especially from Europe who could say other things.
10 But, first of all, with regard to the total
11 protocol design, as the FDA has pointed out in
12 their document as well, the design requirements are
13 not as detailed and specific and uniform as they
14 are now with the international harmonization
15 guidelines.
16 So, if the European studies were to be
17 done today, there would be very detailed analytical
18 statistical analysis plans included in the
19 protocol.
20 I believe, as Dr. Chabac also pointed out,
21 that the purposes of the studies globally were for
22 registration purposes so there was a common plan
23 for analysis of the data and that is why these
24 variables, all of which we consider to be related
25 because they are all another way of looking at
101
1 abstinence or drinking that these are appropriate.
2 DR. CHABAC: I just want to add something.
3 Remember that we designed all our European studies
4 using a core protocol. That means the same study
5 design. In our protocol, we specified which were
6 our primary criteria, mainly time to first relapse
7 since we were seeking for an indication to maintain
8 long-term abstinence. So it was our primary
9 criteria very well described in our protocol and it
10 is in the NDA.
11 DR. LEON: Can I just follow up? In my
12 reading of the documents, it looked like the time
13 to alcohol was not specified as a primary dependent
14 variable either in Pelc or in the second one,
15 Paille.
16 DR. GOODMAN: Right; I think that is
17 correct. They varied slightly between the studies
18 but what I am saying is that the information that
19 was obtained allowed one to do an integrated type
20 of analysis where you could use the information and
21 look at it for a similar outcome parameter. As I
22 said earlier, and I think we all agree, there is
23 not really a methodology, a statistical
24 methodology, people certainly agree on but the
25 outcome measures for this, especially when Lipha
102
1 was working with acamprosate, naltrexone was not
2 available in Europe and just became available there
3 recently. So Lipha was really pioneering this area
4 and the types of outcome parameters that were used
5 were, by that very nature, something that could be
6 gleaned from the information gathered.
7 I think probably each country had their
8 own kind of slant on what they thought, more or
9 less investigator-driven types of endpoints.
10 DR. G. COOK: The primary objective,
11 relatively clearly abstinence.
12 DR. GOODMAN: Yes; exactly.
13 DR. G. COOK: So, even though there may
14 have been variations on how abstinence was looked
15 at, whether it was time to first drink or complete
16 abstinence or number of abstinent days, the focus
17 was on abstinence and the conclusions across those
18 multiple criteria were pretty much the same.
19 I think the analyses the FDA has done
20 pretty much agrees with that so there are not
21 really any major inconsistencies that I have seen
22 if you basically say the real objective of those
23 studies was abstinence.
24 DR. LEON: But I still haven't heard you
25 say that the data analyses that were presented
103
1 today corresponded directly with that that was
2 described before the data were collected. It seems
3 like the primary efficacy measure and the data
4 analytic techniques in all four of the studies are
5 different than those specified in the protocols.
6 DR. G. COOK: But, for the three European
7 studies, your earlier point, which is consistency
8 of findings across a variety of ways of looking at
9 the data, was, indeed, supported. Now, the U.S.
10 study is going to be a totally different phenomenon
11 which we will get to shortly.
12 Essentially, the structure of the European
13 studies, particularly at the time they were done,
14 had a reasonably clear objective of abstinence and
15 the criteria that were looked at were all criteria
16 that were relevant to abstinence. The conclusions
17 across those criteria by the different ways of
18 looking at them, whether by the sponsor or the
19 agency, were pretty much the same.
20 It would be important that they were the
21 same because if it had turned out that the analyses
22 of abstinence in the Europeans had varied according
23 to measure or method, that would be an issue with
24 respect to the European studies. So, the fact that
25 there is consistency across those different ways of
104
1 looking at the data, even though they have
2 different conventions for how you deal with
3 intervals between visits, is important to the
4 robustness.
5 With respect to the U.S. study, I think
6 there was, at one time, interest in the time to
7 first drink or abstinence. That was a goal of the
8 U.S. study. But that was basically defeated
9 because the patients weren't abstinent at baseline.
10 In other words, unlike the European studies, you
11 did not have abstinent patients at baseline. So
12 the notion of looking at time to first drink or
13 total abstinence broke down. That is why other
14 things had to be looked at.
15 Now, the role of the U.S. study here is to
16 try to understand consistency; is there information
17 in the U.S. study that more or less fits with what
18 was proven in the European studies. The U.S. study
19 doesn't prove anything. It is possibly
20 inconclusive. It possibly raises doubt about what
21 was seen in the European studies.
22 So the role of all of the explanatory
23 analyses--we don't call them confirmatory anymore
24 for the U.S. study; we call them explanatory--is to
25 try to understand whether there is information or
105
1 trends in the U.S. study that fits with what was
2 proven in the European studies. That is what Dr.
3 Mason tried to share with you all.
4 So the original planned analyses didn't
5 work because we didn't have an abstinent population
6 at baseline.
7 DR. OREN: Dr. Hamer?
8 DR. HAMER: Actually, I have a related
9 question. I have very little experience in
10 substance abuse but I do have a great deal of
11 experience in depression studies and schizophrenia
12 studies and a variety of other psychiatric studies.
13 If a sponsor came in with four depression
14 studies of which three were positive and one
15 wasn't, basically, I think both the FDA and this
16 committee would tend to sort of shrug our shoulders
17 and say, you know, we have failures in depression
18 studies. Three out of four is not bad. Sounds
19 like a good drug to me.
20 So, as a statistician, I never want to
21 underestimate the pure properties of randomness.
22 So I may not feel as compelled as the FDA seems to
23 feel to seek explanatory reasons for why the U.S.
24 study, unfortunately, failed.
25 Now, there are other issues with the
106
1 European studies having to do with the time frame
2 and conditions under which they were designed and
3 the fact that they didn't have this rigid
4 prespecified endpoints and analyses as the ones we
5 would design now are.
6 But I do agree with Dr. Cook that what we
7 really should be pulling out of the
8 nonprotocol-specified reanalyses of the U.S. data
9 is that these analyses are possibly explanatory.
10 They are hypothesis-generating. They are not
11 hypothesis-confirming. I hope that the sponsor is
12 not claiming that these hypotheses in the U.S.
13 study indeed confirm that acamprosate promotes
14 abstinence in patients who are already abstinent
15 and I would hope we don't interpret it that way.
16 So I would say that our task, in some
17 sense, is, using the standards that we are
18 accustomed to using, in a sense, to look at the
19 European studies and decide whether those provide
20 sufficient evidence of safety and efficacy.
21 DR. OREN: If I could just ask you, since
22 we just have a few more minutes for this
23 segment--we will have an afternoon discussion
24 section to weigh all the different points. So if
25 we could just focus on the specific questions for
107
1 the company to answer.
2 DR. HAMER: In that case, I will postpone
3 things.
4 DR. OREN: Okay. Dr. Keck?
5 DR. KECK: This is a belated follow up to
6 Dr. O'Brien's point about psychosocial influence on
7 outcome. I am just, again, trying to understand
8 the many reasons why the U.S. study failed. In a
9 way, it doesn't surprise me that a study in which
10 you had ambivalently motivated people many of whom
11 were not abstinent to participate in the trial with
12 poly drug abuse didn't do so well in this study.
13 But one other embedded reason I wonder
14 about in the design is it seems to me that patients
15 had not only one but potentially two psychosocial
16 treatments here because of the--I'm getting the
17 terminology here--the time-line follow-back method
18 which, again, coming not as a substance-abuse
19 researcher but doing research in other
20 impulse-control disorders, any time you put a diary
21 into a study as a treatment-outcome measure, you
22 invariably introduce, I think, subtly, a form of
23 behavioral therapy by completion of the diary,
24 itself.
25 So I guess I am saying it seems to me you
108
1 had two psychosocial interventions or behavioral
2 therapy interventions which I think made it even
3 more difficult to find a drug-placebo difference.
4 Does that sound fair to say?
5 DR. MASON: It sounds quite fair and
6 accurate, and the placebo response rate was high in
7 the U.S. study. I completely agree with you that
8 the data-collection methods, in themselves,
9 probably raised the threshold of what was perceived
10 by the patient as therapeutic activity, in addition
11 to the twenty minutes that they were officially
12 assigned.
13 DR. OREN: Dr. Hughes
14 DR. HUGHES: I wonder of you could respond
15 to my rationale here. The notion is that, with
16 increased psychosocial treatment, you decrease the
17 odds ratio between active and placebo. That is the
18 notion I hear being proposed.
19 If increased psychosocial is--the typical
20 way you test that is you take the response of the
21 placebo group and does it correlate with the odds
22 ratio. It is a standard metaanalytic treatment.
23 So the notion is studies that have high placebo
24 responses should have low odds ratios.
25 I did this before I came down. When I
109
1 look across the fourteen studies, that is not the
2 case in the fourteen acamprosate studies. So my
3 rationale is the data don't suggest that high
4 placebo rates lead to lower odds ratios. But maybe
5 I am thinking wrong.
6 DR. GOODMAN: I am certainly far from a
7 statistician but I would just comment that, if you
8 are looking across the European studies and, if I
9 understood your comments correctly, you were
10 talking about behavioral therapy and I gathered
11 something rather substantial that was, as Dr. Mann
12 has pointed out, that was not the case in Europe.
13 It was not consistent and it varied and it was
14 more--the term that was used in the European
15 dossier was "naturalistic."
16 Maybe I didn't understand what you were
17 saying.
18 DR. G. COOK: This is Gary Cook, again. I
19 am not sure how to answer your question. I think
20 when the placebo rate is higher, that can make it
21 more difficult to show a difference in rates
22 because the amount of room for change may be
23 affected.
24 Odds ratios are complicated kinds of
25 things, so their ability to be large or small is
110
1 related to the base rate that you are working with
2 so an odds ratio of 90 percent versus 95 percent is
3 2. If you have 90 percent compared to 95 percent,
4 the odds ratio there is about 2 whereas if you are
5 comparing 50 percent to 67 percent, the odds ratio
6 is 2.
7 So I think it is very difficult to try to
8 actually project what you think an odds ratio might
9 do as you change the base rate. If you do have
10 high placebo rates, it may make it more difficult
11 to show a substantial difference in response rates
12 because the amount of room for improvement may be
13 less.
14 But I think, really, it is uncertain in
15 these kinds of things. Also, again, the U.S.
16 population and European populations were different
17 from one another, so extrapolating across the two
18 populations will have its difficulties.
19 DR. OREN: To conclude this segment, Dr.
20 Rudorfer and then I will ask one question after
21 that.
22 DR. RUDORFER: Thanks. It certainly can
23 be challenging to do an effectiveness study such as
24 the U.S. study where one broadens inclusion
25 criteria to try to better reflect real-world
111
1 populations. I think the American investigators
2 did a very good job of responding to the FDA
3 request to, say, have a broad age range and include
4 cormorbidities.
5 But what concerns me is they are sort of
6 going back to basics. If we are discussing the
7 efficacy of a drug for "maintenance of long-term
8 abstinence from alcohol," I still don't understand
9 why abstinence was not an inclusion criterion.
10 DR. GOODMAN: I will let Barbara address
11 that, but I think our assumption in designing the
12 protocol was that patients would understand that
13 they were to be abstinent at the study onset. It
14 was not explicitly stated, but that was our
15 expectation. So, of course, it was quite a
16 surprise to find out that half these people were
17 not abstinent.
18 I think we had been quite--what would I
19 say--just really tuned into the European
20 populations as starting from this abstinence
21 without appreciating that that would not be the
22 case in our study.
23 But, Barbara, you might--we also had the
24 steady-states idea.
25 DR. MASON: Your point is well taken. The
112
1 behavioral therapy was abstinence oriented,
2 complete abstinence. The admission criteria was
3 people had to have a minimum period of time with no
4 hazardous drinking, which is no more than one drink
5 a day for women, two drinks a day for men, so that
6 they would have decreased to that level so we
7 wouldn't have to deal with withdrawal symptoms on
8 study.
9 But, because acamprosate takes the time
10 that it does to reach steady state, and the animal
11 literature was indicating that there may be some
12 benefit in alcohol withdrawal, our idea was to
13 start drug as soon as possible in the process to
14 help these patients become and stay abstinent.
15 That is why the admission criteria were
16 what they were. We did no interim analyses or
17 peaks or anything and so that is why it was the
18 surprise that it was in terms of the rate of
19 nonabstinence.
20 DR. OREN: My question is for Dr. Mann.
21 In the European studies in support of the efficacy
22 of acamprosate, you mentioned that the completor
23 rates were higher in the active group than in the
24 placebo group. Was that a statistically
25 significant difference and what kind of statistic
113
1 was used?
2 DR. MANN: It was a significant
3 difference. I think we could pop up one of these
4 extra slides, but I know there was a significant
5 difference between those two groups but I don't
6 recall what kind of statistics we did. But we
7 could find out and then deliver that information
8 later if you want. Sure.
9 DR. OREN: We will now take a ten-minute
10 break and then reconvene to hear from the FDA.
11 [Break.]
12 DR. OREN: We are now at the point for FDA
13 presentations. I will call upon Dr. Celia
14 Winchell, Medical Team Leader for Addiction Drug
15 Products.
16 FDA Presentations
17 Clinical Issues on Efficacy
18 DR. WINCHELL: I am Celia Winchell from
19 the FDA and I am going to speak to you this morning
20 about the clinical review of the efficacy of
21 acamprosate.
22 [Slide.]
23 I want to let you know that we approached
24 this data hopefully. We knew before the
25 application came in that the American trial hadn't
114
1 worked out. But it isn't unusual for an
2 application to contain some trials that worked and
3 some trials that weren't able to show a difference
4 from placebo.
5 But this time, we had some older, perhaps
6 less rigorous, foreign studies that worked against
7 a recent domestic and really good study that
8 didn't. It was hard to overlook that.
9 We had some reservations about the conduct
10 of the European trials but we looked at them at
11 them a few different ways and we were able to find
12 encouraging results. Then both the statistical
13 reviewer, Dr. Wang, and I dug into the American
14 trial data. We really hoped there would be some
15 explanation for the outcome that would have some
16 face validity and could tell us something about
17 circumstances in which acamprosate works and
18 circumstances in which it doesn't and that would
19 give us confidence that we could accept the
20 European studies.
21 For about the next half hour, I am going
22 to take you through the process of looking at the
23 efficacy data and show you where it led us.
24 [Slide.]
25 The questions on this slide are the ones
115
1 you have been asked to consider this morning. You
2 have heard some comments from Lipha on the matter
3 and, before I begin, I will point out that there
4 are two ways of casting the questions.
5 It was suggested in the materials that we
6 reviewed that the reason the European trials were
7 able to demonstrate the effect of acamprosate and
8 the American trial wasn't is primarily that the
9 populations differed. The European subjects, as we
10 have heard, all randomized to treatment after
11 completing an inpatient detox. There were few
12 polysubtance abusers in the European studies and
13 the European studies either assumed or required a
14 high level of motivation for abstinence.
15 Lipha was able to identify a subset of the
16 American population they presented to us as being
17 most like the European subjects and they feel that
18 this group did demonstrate the effect of
19 acamprosate. So you could put the questions on the
20 slide this way. Given the positive findings
21 throughout Europe, how would we weigh the results
22 of the United States trial upon consideration of
23 our explanatory analyses based on population
24 differences?
25 But, on the other hand, I found a number
116
1 of the aspects of the European data presentation
2 that gave me pause during my review and I was
3 completely unable to find a way to explain the
4 results of the American trial.
5 As you saw in the materials provided in
6 the backgrounder, I defined a number of population
7 subsets that I thought could account for the
8 differences. For statistical reasons, I restricted
9 myself to use of prerandomization characteristics
10 and, no matter how I sliced it, there was no
11 treatment effect of acamprosate at the proposed
12 dose. It was not a matter of failure to reach
13 statistical significance due to small sample size.
14 There was really no difference and occasionally
15 there were differences that trended in the wrong
16 direction, in the direction of favoring placebo.
17 So I would be inclined to put the
18 questions this way. In view of the failure of the
19 carefully conducted American trial, which we are
20 unable to explain through analyses directed at
21 various subpopulations, can we accept the findings
22 of the European studies knowing the data was
23 collected less systematically.
24 [Slide.]
25 In the next few minutes, I am going to
117
1 take you through the review of the efficacy data
2 that was submitted to the FDA for review. The
3 emphasis in the material submitted to us for the
4 purpose of an integrated efficacy analysis was on
5 the cumulative abstinence duration. So I will
6 cover how we concluded that this outcome variable
7 identified for the European pivotal trials couldn't
8 really be viewed with confidence.
9 Then I will give you the good news about
10 what we were able to make out of those trials and
11 then I will walk you through the American trial
12 which wasn't able to show an effect of acamprosate
13 and our attempts to resolve the discrepancies
14 between these bodies of data.
15 [Slide.]
16 So, first, what is my problem with
17 cumulative abstinence duration. As I mentioned,
18 the primary outcome variable emphasized in the
19 integrated analyses in the European pivotal trials
20 was cumulative abstinence duration, which is
21 measured in days, or what was called corrected
22 cumulative abstinence duration which amounts to
23 percent days abstinent.
24 In your briefing book, you read that we
25 rejected this variable on review. I will remind
118
1 you that these studies were complete at the time
2 the IND was open. We never discussed the design
3 and analysis of these trials prospectively, so
4 there wasn't an opportunity to comment prior to the
5 NDA review.
6 Let me make the point that I have no
7 problem with these measures in theory. They are
8 attractive because they capture the picture of
9 drinking behavior even for those subjects who don't
10 abstain for the entire observation period which we
11 know is most of them.
12 The problem with the use of these measures
13 in analyzing the European studies is that they
14 amount to a false precision. These studies
15 collected the drinking data in a somewhat
16 nonsystematic way at widely spaced visits and used
17 various data-handling rules to convert the data so
18 collected into number of days of abstinence and
19 days of drinking.
20 On examining the protocols, the
21 case-report-form fields and the data-handling
22 rules, I concluded that the CAD in the three
23 pivotal European trials, actually in all the
24 European trials, the ten additional ones other than
25 the British study, seem to be a highly imputed
119
1 value that went beyond the precision of the data
2 actually collected.
3 I will walk you through the three pivotal
4 studies to show you what I mean.
5 [Slide.]
6 We have already heard about these studies.
7 The first study is Pelc II. This was a short-term
8 study with 90 days of treatment. This study had
9 seven on-treatment visits. These visits were close
10 together, one to two weeks. At each visit, the
11 investigator estimated the subject's average daily
12 consumption on drinking days and average frequency
13 of consumption. It wasn't a systematic approach to
14 this, like the time-line follow-back method.
15 The real problem, though, is in the
16 data-handling rules. Anyone who had any number
17 other than 0 listed for frequency and amount for
18 the purposes of the CAD calculation was considered
19 to have been drinking during the entire inter-visit
20 interval. So any number between one drinking day
21 and 15 drinking days was transformed to 15 drinking
22 days.
23 If a visit was missed, drinking days were
24 imputed all the way back to the previous visit. So
25 this method collapses a fairly wide range of
120
1 responses into two possibilities, 0 or 15. I find
2 this troubling because the result was then
3 mathematically summed, a mean was calculated to the
4 tenth place and comparisons were made
5 statistically.
6 [Slide.]
7 The Paille study, somewhat more
8 problematic, had a one-year treatment period but
9 only nine visits on treatment so the interval was
10 as much as 60 days between visits. At these
11 visits, the investigator again came up with an
12 estimate of the number of days of nonabstinence and
13 the drinks per drinking day without a systematic
14 technique for reconstructing the data.
15 But, unlike the Pelc study which used this
16 very conservative approach, the Paille study
17 handling rules took that estimate on its face and
18 put it into the calculations of CAD. I am just
19 skeptical about the precision.
20 [Slide.]
21 The PRAMA study had only six visits over
22 48 weeks of treatment. For half the visit, the
23 intervisit interval was three months. At these
24 visits, there was a global assessment by the
25 physician and then the physician was also supposed
121
1 to determine if a relapse occurred, classify it as
2 short-term or long term, try to figure out when it
3 happened, and then there were data-handling rules
4 for the calculation of CAD which are so complex
5 that I have put them on a separate slide which is
6 still too small to read.
7 [Slide.]
8 I know you can't read this but I am just
9 trying to make the point that there is such a
10 complicated set of mathematical rules here to
11 transform what is a rough estimate about what has
12 happened for the past three months into a specific
13 number of days of drinking versus abstinence. I
14 just felt that is a false precision that goes
15 beyond what was really known. That was the bad
16 news.
17 [Slide.]
18 But the good news was we looked at the
19 datasets and tried to see what we could conclude
20 based on the data collected. You have heard that a
21 considerable amount of effort went into
22 establishing abstinence versus nonabstinence.
23 There were blood-alcohol levels taken,
24 breathylizers. There were collateral informants.
25 There were other external informants. There was a
122
1 lot of effort here. So we could place some
2 credibility on that.
3 I considered how many people were assessed
4 by the investigator as continuously abstinent. I
5 realize that is a very high standard and doesn't
6 really capture all of the clinical effect that
7 would be considered relevant so I wanted a way
8 other than CAD to look at periods of abstinence
9 even if they were interrupted by periods of
10 drinking.
11 So I went through and I counted how many
12 people had zero visits at which they were assessed
13 as abstinent, how many had two, and so on, and
14 compared across treatment groups. Now, with an
15 intervisit interval of 90 days, binary assessment
16 of abstinence versus nonabstinence, maybe a little
17 suspect but we talked about all the effort that
18 they went to do this; right?
19 If the subject can convince the
20 investigator he hasn't had a drink in three months,
21 that probably does mean something.
22 [Slide.]
23 Here I have laid out the results of the
24 continuous abstinence analysis. This lists the
25 number of percent of subjects in each treatment arm
123
1 who were assessed as continuously abstinent
2 throughout the treatment period. In each of these
3 studies, acamprosate, at the dose proposed for
4 marketing, was superior to placebo and the
5 differences were statistically significant.
6 Here Pelc is clear. Here this one is kind
7 of marginal and it depends on what analysis you do.
8 Mine came out with a p-value of 0.042 for this
9 pairwise comparison. Then here I will just
10 clarify. This says 1998 per day. Actually these
11 patients were allocated by weight so that heavier
12 patients got 1998 per day and later patients got
13 1332 per day. But it turns out there were only
14 thirteen people who got 1332 a day. So, for
15 convenience, I am just calling it 1998 a day and
16 this is also statistically significant.
17 [Slide.]
18 I also wanted to look at the results that
19 included noncontinuous abstinence as clinically
20 relevant. So I tabulated for each study how many
21 subjects were assessed by the investigator as
22 abstinent at zero visits, one visit, two visits and
23 so on. I am going to show the tables for each
24 study one-by-one on the next few slides. I will
25 just tell you that the differences come out
124
1 statistically significant in favor of acamprosate
2 1998 milligrams per day in all the studies.
3 If you look very closely, it seems that,
4 for the most part, the superiority in this analysis
5 continues to be driven primarily by the subjects
6 who were continuously abstinent. But there is, in
7 some studies, a little greater tendency for the
8 placebo subjects to have very few abstinent visits.
9 In other studies, the subjects who have many but
10 not all abstinent visits strengthen the finding.
11 So this first slide shows you the results
12 from Pelc II. There were supposed to be nine
13 visits but, for some reason, there are no subjects
14 with nine abstinent visits. But these numbers
15 here, this 26, 26, 9, these are the same numbers
16 that come up for the continuously abstinent
17 analysis. I haven't been able to explain this. It
18 may have to do with handing of missing data.
19 In any case, you will see here that there
20 is a greater tendency for placebo subjects to have
21 zero, one or two abstinent visits as compared to
22 people assigned to active condition.
23 Of course here you will see this is
24 consistent with the continuous abstinent analysis.
25 There are just a lot more people assigned to active
125
1 who had eight visits compared to placebo.
2 [Slide.]
3 Here is the analysis for the Paille study.
4 There were nine visits, these 85 people here.
5 These are 85 people who were continuously
6 abstinent. They are shown as having nine visits
7 assessed as abstinent.
8 This study was the one that had the most
9 marginal results when you look at continuous
10 abstinence. But you can see that if you add in the
11 people who had eight abstinent visits, that
12 strengthens the finding because you end up with 44
13 in each of these groups which is 24 to 25 percent.
14 In the placebo group, you end up with 22,
15 which is only 12 percent. At the other end of the
16 spectrum, the least successful end, the difference
17 is less obvious. 54 percent of the placebo group
18 has two or fewer abstinent visits--I am adding
19 these together--compared to 47 percent of the
20 acamprosate low-dose group and 40 percent of the
21 acamprosate 1998 milligram group.
22 [Slide.]
23 Here is the data from PRAMA. There were
24 only six visits in this one-year study and, as Dr.
25 Wang will discuss, there were many dropouts and
126
1 dropouts occurred at different rates across
2 treatment groups. Missing visits couldn't be
3 assessed as abstinent visits. They were missing.
4 So this analysis is vulnerable to the dropout
5 problem. We understand that many fewer placebo
6 subjects actually attended six visits so,
7 obviously, they have many fewer opportunities to be
8 assessed as abstinent.
9 So we have to look at this analysis with
10 caution in view of that phenomenon. But here you
11 see that the superiority of acamprosate over
12 placebo at the most successful end of the spectrum
13 is clearly driven by the subjects with six
14 abstinent visits. There is no difference at four
15 or five.
16 But the difference between treatments is
17 also apparent at the other end of the success
18 spectrum. 63 subjects, or 46 percent in placebo
19 group, had zero or one visit at which they were
20 assessed as abstinent as compared to 39 which is
21 just 29 percent of the acamprosate group.
22 [Slide.]
23 So, in summary, it does look as if the
24 three European studies indicate an effect of
25 acamprosate in maintaining abstinence after
127
1 detoxification.
2 [Slide.]
3 Let's turn to the American study. As you
4 have already heard, this was a multicenter study
5 involving 601 subjects at 21 centers throughout the
6 United States, 260 subjects randomized to placebo,
7 258 on acamprosate 2000 milligrams a day and 83 to
8 the exploratory arm, acamprosate 3000 milligrams a
9 day.
10 We have discussed that the study used a
11 different formulation from the one in the European
12 trials. In those studies, there was a 333
13 milligram tablet. Subjects took two tablets three
14 times a day with meals. This study used a
15 compositionally proportional 500 milligram tablet.
16 The subjects actually took three tablets
17 QAM and QHS3. So everybody, including the 3 gram
18 exploratory arm, would have to take three tablets,
19 the 2 gram got two active and one placebo and the
20 placebo arm got three placebo.
21 We have already discussed the
22 pharmacokinetics and the TID dosing isn't essential
23 to maintaining steady state. So BID is not a
24 concern. And we have already touched on the food
25 effect. The effect of food is to lower systemic
128
1 exposure so, if anything, we think that the dosing
2 schedule in the American study would have exposed
3 the American subjects to a higher total daily dose
4 even though the nominal dose, 2 grams and 1998
5 milligrams, are essentially the same.
6 This was a carefully conducted and closely
7 monitored study. The features included six months
8 of treatment with eight on-treatment visits most of
9 which were at four-week intervals. Subjects
10 brought drinking diaries to each visit which were
11 used to help reconstruct day-by-day drinking data
12 using the time-line follow-back method.
13 Breath alcohol was measured at each visit
14 and collateral informant data was also collected at
15 intervals. This information was used to modify the
16 drinking data when it conflicted with the subject's
17 information and, in addition, as we have heard, the
18 subjects received a standardized brief psychosocial
19 therapy oriented to reinforcing medication
20 compliance.
21 [Slide.]
22 The primary outcome measure was the
23 percent of study days which were non-drinking days
24 referred to in the study report as corrected
25 cumulative abstinence duration. The number of
129
1 non-drinking days was calculated from the time-line
2 follow-back data as modified by other information
3 in the breath alcohol collateral informant and
4 there was an algorithm prespecified for assigning
5 values to missing days that occurred prior to
6 discontinuation or lost to follow up.
7 There was also a fairly rigorous protocol
8 for locating subjects to minimize the amount of
9 data that had to be imputed. My understanding is
10 that, in the calculation of a CCAD,
11 discontinuations were evaluated by a blinded panel
12 of raters and, if they were related to drinking,
13 all the days after discontinuation were considered
14 drinking days. But if a discontinuation was not
15 considered related to drinking, the denominator was
16 then adjusted so that the days after dropout were
17 not considered in this calculation of percent days
18 abstinent.
19 You might think that that is very
20 conservative and unfair to people who drop out
21 early as the result of drinking, so we actually
22 looked at people's baseline level of drinking to
23 see, if they got worse, maybe they would go back to
24 how bad they were before they came into the trial.
25 It does probably overestimate but over
130
1 half the subjects were drinking six or seven days a
2 week. About a quarter of them were drinking four
3 or fewer days a week. So it is an overestimate but
4 it is not horrendous.
5 [Slide.]
6 These are the results that I get from the
7 sponsor's datasets that were submitted to us for
8 review. Considering the entire intent-to-treat
9 population, the mean percent days abstinent for the
10 placebo group was 51 percent. The small group that
11 was randomized to 3 grams a day, about the same, at
12 50 percent. And the group that got the recommended
13 dose of acamprosate 2 grams a day had a mean
14 percent days abstinent of 46 percent.
15 Looking at the medians, placebo also
16 outperformed acamprosate. Why did this happen? If
17 acamprosate worked in the European studies why
18 didn't it seem to work here?
19 [Slide.]
20 Here were the simplest and most attractive
21 explanations presented to us from even before the
22 NDA was submitted. First, the European subjects
23 had been detoxed and were abstinent at baseline but
24 the American subjects were not required to undergo
25 detox probably as a consequence of the current
131
1 climate in our medical-care delivery system. Only
2 about 10 percent of them got it.
3 Furthermore, by the time the
4 study-medication treatment began, about half the
5 subjects were already actively drinking. So, the
6 first idea that springs to mind to all of you is
7 that acamprosate is just a relapse-prevention
8 agent. It keeps alcoholics from taking the first
9 drink but it can't seem to put the brakes on if
10 someone is actively drinking.
11 So, of course, I looked at the subset that
12 was abstinent at baseline which is about half the
13 subjects.
14 Now, the second difference was level of
15 motivation. Some of the European studies actually
16 required, as a condition of entry, that the subject
17 be committed to abstinence. Others didn't, but it
18 has been assumed the subjects must have been
19 motivated because they were willing to go through
20 detox.
21 Now, I am not sure about that because I
22 don't know about the healthcare delivery system in
23 Europe, either now or at the time these studies
24 were done over ten years ago. It is possible
25 inpatient detox was pretty standard and readily
132
1 available and that willingness to go into the
2 hospital for three days wasn't really a marker for
3 a high level of motivation.
4 But let's say it was. In the American
5 study, as you heard, subjects were asked to
6 indicate at screening what their goal was for
7 treatment and they could choose from a list that
8 ranged from total abstinence to no goal. You saw
9 that it included temporary abstinence, controlled
10 drinking. You also saw there was another option on
11 there; total abstinence, but I realize a slip is
12 possible.
13 This wasn't, "I think a slip is okay," or,
14 "My therapist has told me, you will probably slip."
15 That's okay. Let's talk about what we are going to
16 do about it. This was just, my goal is total
17 abstinence but I realize a slip is possible. It
18 was multiple choice.
19 I regard that as just as motivated but a
20 little more realistic. And I put those two
21 together. That is actually 72 percent of the
22 subjects and evenly distributed once you add them
23 together, evenly distributed across treatment arms.
24 Finally, the high rate of polysubstance
25 abuse in the American trial was striking,
133
1 especially given that a positive urine tox for
2 anything other than marijuana was exclusionary.
3 Now, only PRAMA of the European studies gave us
4 information about other substance-abuse history
5 and, in that study, on 20 percent of the subjects
6 had any history of other substance abuse.
7 In contrast, the United States population,
8 only 20 percent did not have a history of illicit
9 drug use.
10 [Slide.]
11 As you have heard, Lipha was able to find
12 a subset they thought resembled the European
13 population. It was defined by some
14 post-randomization variables, post-randomization
15 compliance with visits and medication. In
16 addition, a treatment goal of complete abstinence.
17 In this group, the acamprosate arm had 70
18 percent days abstinent and the placebo group had 63
19 percent. But the problem here is that it appears
20 to be that this is the only population that
21 demonstrates an effect of acamprosate. It is
22 defined primarily by post-randomization behavior
23 such as medication compliance and observed use of
24 substances. All post hoc analyses make us
25 uncomfortable because if you do enough of them, you
134
1 are bound to find one that comes out significant
2 which actually makes it particularly troubling that
3 we couldn't.
4 But subset analyses, whether post hoc or
5 planned, that rely on groups defined by
6 post-randomization factors are particularly
7 troubling. Finding that a drug was particularly
8 effective in a group with a certain set of
9 post-randomization behaviors really doesn't give us
10 any information that we can use for patient
11 selection.
12 What's more, this population definition
13 doesn't even take into account the issue of
14 abstinence at baseline which the proposed label
15 indication now indicates is the important feature
16 of patient selection. So I am not convinced by
17 this finding. I am not convinced by this
18 population definition.
19 As you read, I conducted a series of
20 subset analyses of my own using populations that
21 seemed to make sense to me.
22 [Slide.]
23 I am going to go over for you my analysis
24 populations and how I hit upon them. I analyzed a
25 subset of subjects that were abstinent for at least
135
1 five days at baseline. That is fairly
2 straightforward. The subset that identified a goal
3 of abstinence, whether or not they indicated that
4 they realized a slip was possible. And I tried to
5 figure out the best way to define the
6 nonpolysubstance-abusing population. So let me
7 take you through some of the things I considered.
8 First, there was something called an
9 illicit drug use index calculated for each subject.
10 If they had no history whatsoever of illicit drug
11 use, that was zero. So I looked at that group, but
12 it was very, very small. It was 20 percent of the
13 randomized population.
14 So then I thought, well, maybe past-year
15 drug use was probably a reasonable indicator of
16 current active polysubstance abuse. So I looked at
17 the group with no illicit drugs in the past year
18 which enlarged the subset to about 40 percent of
19 the randomized population.
20 But, because subjects were allowed to
21 enter the study if they had a tox screen positive
22 for marijuana, if I looked at the group that had no
23 past-year drug use other than marijuana, I actually
24 got as many as 80 percent of the randomized
25 population. Now, I will acknowledge that that is
136
1 our fault. We asked Lipha to broaden the inclusion
2 criteria to allow for a positive tox for marijuana
3 at entry because we are concerned that the actual
4 target population has a pretty high prevalence of
5 polysubstance abuse and it seems like we were
6 right.
7 Even though people were screened out, if
8 they had current dependence on any other substance
9 and screened out if they had a positive urine tox
10 at screening for anything other than marijuana, the
11 enrolled population still has a 14 percent history
12 of opiate use and 49 percent history of cocaine
13 use. This is what American alcoholics look like.
14 I also looked at the group defined by the
15 results of urine toxes during the study. But I am
16 actually not at all convinced that this is useful.
17 With monthly study visits, tox screens are unlikely
18 to pick up all the illicit drug use in the study
19 and also nothing can be predicted about the results
20 of urine-tox screens that weren't done because the
21 subject dropped out of the study.
22 So if you select subjects who just don't
23 have urine-tox evidence of drug use, it doesn't
24 mean you have a population that didn't use drugs.
25 It also especially means you don't have a
137
1 population that is prone to use drugs after they
2 drop out. Also, there were only urine-tox data for
3 525 subjects, so I didn't use this.
4 Ultimately, I decided to focus on the
5 subjects whose only illicit drug use in the past
6 year had been marijuana. So, from now on when I
7 say no past-year illicit drug use, what I am
8 talking about is actually the subjects who had no
9 past-year illicit drug use other than marijuana.
10 Then I put together the subset that was
11 abstinent at baseline, motivated and had no
12 past-year illicit drug use, a very small group,
13 only 20 percent of the randomized population.
14 [Slide.]
15 These are my results. This is using the
16 sponsor's corrected cumulative abstinence duration
17 in the dataset. Here is motivated. I don't have a
18 slide for this but I did look. I looked at
19 motivated, total abstinence versus total
20 abstinence, but I believe a slip is possible. And
21 they are exactly the same. They are the same.
22 Here is abstinence. Here is the no
23 history whatsoever of drug use. These are very
24 small numbers. No illicit drugs and here is not
25 illicit drugs other than marijuana. This is the
138
1 last time you are going to see these guys. You
2 will see that these are all actually going the
3 wrong way.
4 I have to say that, going into this, I was
5 really hoping that the rubber was going to meet the
6 road somewhere. I was going to be able to say,
7 ah-ha, it only works in pure alcoholics, or, see,
8 as long as you are abstinent at baseline, it works.
9 But these analyses just don't bear out any
10 conclusion about patient selection that suggests
11 why acamprosate didn't work in this study.
12 [Slide.]
13 Looking at the subset that was abstinent
14 and motivated and the subset that was abstinent,
15 motivated and had no past-year illicit drug use, I
16 still could not find an effect of acamprosate.
17 [Slide.]
18 I looked at other measures, too. Complete
19 abstinence wasn't very useful because there were so
20 few subjects, 33 to be exact, who were abstinent
21 for the entire trial and 20 of them were on
22 placebo.
23 There was a categorical analysis of good
24 response which looked at how many subjects had 90
25 percent days abstinent or more. This was
139
1 interesting because the motivation ITT population
2 defined by the sponsor did show the acamprosate
3 group tied with the placebo group and then the
4 sponsor defined motivated efficacy evaluable
5 population showed acamprosate beating placebo, but,
6 as it turned out, my analysis populations do not
7 fare as well and the placebo group did better than
8 the acamprosate group in all the populations that I
9 tried.
10 [Slide.]
11 So next I looked at fairly liberal
12 definition of success. There was a dataset in
13 which relapse was flagged if the patient relapsed
14 into having at least five drinks a day for five of
15 the next seven days.
16 So we looked at how many subjects never
17 had a relapse as so defined. Obviously, success by
18 this criterion is fairly common. In this slide,
19 you will see the ITT population looks a little bit
20 promising but neither the abstinent subset, the
21 motivated subject, the no-past-year-illicit drugs
22 or the group that met all three criteria show an
23 effect of acamprosate on this measure. But the
24 sponsor motivated efficacy evaluable does.
25 [Slide.]
140
1 Just in case I missed something, I pored
2 over the demographics from the different trials to
3 find another explanation. I was so enthusiastic
4 about this that I misinterpreted this data and I
5 confused the number of drinks per drinking day with
6 the average number of drinks per week and I was
7 under the misimpression that there were more heavy
8 drinkers in the European data.
9 But, just in case you were wondering, this
10 analysis doesn't work either.
11 [Slide.]
12 In summary, the European studies indicate
13 an effect of acamprosate on either a continuous
14 abstinence or noncontinuous abstinence while the
15 American study does not demonstrate the efficacy of
16 acamprosate in any subset defined by
17 prerandomization variables that would be useful for
18 patient selection.
19 [Slide.]
20 So I will put the questions back up here.
21 I have gone through some of the concerns about the
22 data from the European trials; relatively
23 nonsystematic data collection, low frequency of
24 study visits and then some of the ways in which the
25 European-trial populations differed from the
141
1 American population.
2 Then I went through the exploratory
3 analyses I undertook to try to select the subgroup
4 from the American study who resembled the European
5 population on important measures such as level of
6 motivation, baseline drinking status and
7 polysubstance abuse and I showed you that I was not
8 able to identify any population that demonstrated
9 the effect of acamprosate on measures including
10 percent days abstinent, categorical good response,
11 or even the fairly low bar of surviving the trial
12 without five heavy drinking days in a single week.
13 So I will reiterate my way of looking at
14 the questions we have posed to you. In view of the
15 failure of the carefully conducted American trial
16 which we were unable to explain through analyses
17 directed at various subpopulations, can we accept
18 the findings from the European studies knowing that
19 the data was collected less systematically?
20 I am going to turn the microphone over to
21 Dr. Sue Jane Wang for the statistical presentation.
22 Statistical Perspective of Acamprosate Experience
23 DR. WANG: Good morning, everyone. I am
24 Sue Jane Wang from Statistical Discipline of FDA.
25 [Slide.]
142
1 In this presentation, I would focus on the
2 statistical perspective of my acamprosate review
3 experience. [Slide.]
4 Here is the outline of today's
5 presentation. First, I will discuss the dropout
6 issue in the three European trials followed by
7 proper interpretation of the efficacy results. I
8 will spend most of the time on the U.S. trial
9 because the drinking data was much more credible in
10 this well-controlled study but knowing that the
11 differential dropout problem still exists in the
12 U.S. trial making it very difficult to interpret.
13 Finally, I would bring to your attention
14 on the conflicting analytical issues we faced
15 during review in the U.S. trial and the European
16 trials.
17 [Slide.]
18 Since you have heard several
19 presentations, I will just use the following
20 notations for the four dose arms that consist of
21 these four different studies: first, the placebo
22 arm; acamprosate, low dose, only studied in the
23 European; acamprosate, medium dose studied in both
24 different places; and the high dose, 3 grams per
25 day.
143
1 [Slide.]
2 The Pelc trial was a multicenter
3 double-blind, randomized, placebo-controlled
4 three-arm study. The objective of this study was
5 to explore the effectiveness and tolerance of
6 acamprosate in helping to maintain abstinence in
7 the weaned alcoholic patient population. Although
8 the main criteria of judgment was the consumption
9 of alcohol, the drinking data was based on
10 respective collections from clinicians. The Pelc
11 II study was the shortest, about three months study
12 duration.
13 [Slide.]
14 The number of patients in this study was
15 about 60 for each treatment arm. Among this
16 percent of patients who discontinued study early
17 was the highest with placebo, 48 percent, and lower
18 but similar for the low-dose and medium-dose
19 acamprosate, about 30 percent. Time to
20 discontinuation from the study was similar among
21 the three groups.
22 To analyze the percent of patients with no
23 relapse, two analysis results are presented. Let
24 me explain the two analyses first. For the dropout
25 of this analysis, patients who did not complete the
144
1 study and did not relapse will be considered as a
2 good outcome or a success. So the numerator is the
3 number of patients who did not relapse but who may
4 or may not complete the study.
5 This is the traditional
6 last-value-carried-forward analysis. Often, the
7 additional trial considers dropout patients as a
8 bad outcome. However, in light of very different
9 dropout patterns between the U.S. and the European
10 trials, we think it is important to show these
11 analysis results.
12 The other one, see the row of as relapsed.
13 Only patients who completed the study and did not
14 relapse is considered as a good outcome. Although
15 a patient may discontinue the study and did not
16 have any relapse at the time of discontinuation,
17 but in this analysis they would be considered as
18 relapsed.
19 As shown in this table,
20 acamprosate-treated patients had more than twice on
21 the percent of no relapse as compared to placebo
22 using either the dropout-as-is analysis or the
23 as-relapsed analysis. In addition, the finding of
24 the time to first relapse was consistent with the
25 percent of no-relapse rates. All showed convincing
145
1 evidence of acamprosate effect.
2 [Slide.]
3 The Paille was a multicenter double-blind,
4 randomized, placebo-controlled study with three
5 arms. Although the low dose and medium dose were
6 included in this trial, the main objective was
7 really to study the low dose not the medium dose in
8 the alcohol patients who were followed as
9 outpatients after withdrawal.
10 In this 360-day trial, patient size was
11 about 180 per arm. Similar to the Pelc II trial,
12 significantly more dropouts occurred in the
13 placebo-treated patients compared to the two
14 acamprosate groups, 65 percent versus 55 and 48
15 percent. But the treatment exposure time was the
16 shortest with the placebo, about eight months,
17 followed by the low-dose acamprosate of 10.5 months
18 and the high dose, 11.8 months.
19 When the LVCF type analysis was
20 performed--that is, the dropout-as-is
21 analysis--there was no statistically significant
22 percent of complete abstinence between acamprosate
23 and placebo although a numerical trend was
24 observed, 23 percent in placebo, 27 in acamprosate
25 low dose and 20 percent in the high dose. The
146
1 p-value was 0.285, not significant.
2 In contrast, when the as-relapsed analysis
3 was performed, twice higher in the percent of
4 complete abstinence was observed with acamprosate
5 as compared to placebo with a nominal p-value of
6 0.044. Interestingly, the sponsor reported that
7 the percent of complete abstinence using 340 days
8 as the cutoff instead of 360 days of the trial
9 period possibly related to the visit window in
10 counting the number of days.
11 [Slide.]
12 In this analysis, as you can imagine, it
13 lies between the dropout-as-is analysis and the
14 as-relapsed analysis giving a nominal p-value
15 somewhere in between, in this case, 0.096, not
16 significant.
17 A closer look using the time to first
18 relapse outcome showed that the time to first
19 relapse was twice longer with the medium dose but
20 not the low dose when compared to placebo, two
21 months versus one month. It is noted again that
22 the trial objectively planned to study the low dose
23 but not the medium dose. Thus the low-dose effect
24 cannot be conclusively shown and the medium-dose
25 effect observed was exploratory but was consistent
147
1 with the Pelc II trial.
2 [Slide.]
3 This is the third study for the European
4 trials. The PRAMA trial was a 48-week multicenter
5 double-blind randomized placebo-controlled two-arm
6 study studying acamprosate versus placebo. The
7 objective here again is to help maintain abstinence
8 after detoxification in the alcoholic patient
9 population.
10 [Slide.]
11 I would like to point out here that the
12 primary efficacy outcome for this study was
13 prespecified and that was time to first relapse.
14 Here, the relapse included short-term relapse,
15 long-term relapse and continuous relapse.
16 [Slide.]
17 In this 48-week trial comparing
18 acamprosate versus placebo, there were 136 patients
19 per group. Again, significantly higher dropout
20 rates were observed in placebo, 60 percent, versus
21 42 percent in acamprosate and had about half the
22 time on trial. It appeared that more placebo
23 patients dropped out because of patient refusal.
24 The percent of abstinence was higher in
25 acamprosate, 51 percent, versus 40 percent in
148
1 placebo when using the dropout-as-is approach. The
2 rates were significantly smaller using the
3 as-relapsed approach as this is more conservative,
4 29 percent in acamprosate and 12 percent in
5 placebo. Note that, in this study, the primary
6 efficacy endpoint prespecified was the time to
7 first relapse. Using the sensory indictor based on
8 either dropout-as-is or as-relapsed, the results,
9 based on time to first relapse clearly showed a
10 significant acamprosate effect.
11 [Slide.]
12 In summary, the three European trials had
13 the drinking data retrospectively collected and the
14 dropout rates were higher in placebo than in drug.
15 The effect of the medium dose was shown in percent
16 complete abstinence in Pelc II, in PRAMA,
17 confirmatory. In Paille, though, exploratory.
18 By the way, the medium dose is the
19 sponsor's proposed to-be-marketed dose. The effect
20 of the low-dose acamprosate was not shown in the
21 Paille trial. I would like to point out that these
22 trials were planned and conducted in the late '80s
23 and early '90s, about a decade ago.
24 [Slide.]
25 Now I would like to turn to the U.S.
149
1 trial. Subjects who were alcohol-dependent or who
2 had been withdrawn from alcohol or who had
3 completed medicated detoxification within two to
4 ten days of study entry were studied. This was a
5 multicenter, double-blind, randomized,
6 placebo-controlled study.
7 I would like to point out that the
8 randomized allocations of patients to the three
9 treatment arms were well balanced. The alcohol
10 measurements were rigorously collected according to
11 alcohol time-line follow-back schedule.
12 [Slide.]
13 For the U.S. trial, the primary objective
14 was to confirm the safety and efficacy of this
15 medium-dose acamprosate. The secondary objective
16 was to explore the efficacy and safety of the high
17 dose. The exploration was only planned for
18 one-third of the patients; that is, 83 patients
19 compared to 260 patients of the other two treatment
20 groups.
21 The treatment phase was 24 weeks or six
22 months and was conducted much more recently,
23 between '97 and '99.
24 [Slide.]
25 There was an apparent difference in the
150
1 percent of patients who dropped out of the study
2 early. Noticeably, the medium dose, or, say, the
3 to-be-marketed dose proposed by the sponsor,
4 appeared to have about 60 percent of patients who
5 discontinued study early but less so in the other
6 two arms, 45 percent placebo, 48 percent in the
7 high dose. The difference was primarily that the
8 medium-dose acamprosate group had more patients
9 dropped out due to patient decision, due to
10 patients lost to follow up.
11 There was also a difference in the time to
12 treatment discontinuation, about one month shorter
13 in the medium-dose acamprosate compared to the
14 other two arms.
15 [Slide.]
16 Here are the protocols specified by
17 primary efficacy outcomes that you are now familiar
18 with.
19 [Slide.]
20 Here are the results of the five primary
21 efficacy endpoints extracted from the sponsor's NDA
22 report and confirmed by us. For the comparison
23 between the medium-dose acamprosate and the
24 placebo--that is, the main objective--the percent
25 of patients who relapsed to drinking were similar,
151
1 92 percent with medium-dose acamprosate and 89
2 percent with placebo.
3 The median time to first drink was four
4 days in both groups and the median time to first
5 heavy drinking days was only a two-day difference.
6 For these three outcomes, the p-value were between
7 0.85 to 0.9 as for the cumulative abstinence
8 duration outcome, or the percent of cumulative
9 abstinence duration.
10 I would like to make a point of this
11 notation here that the sponsor used because I will
12 be referring to that later. CAD, cumulative
13 abstinence duration, in days; CCAD, percent of days
14 abstinence--in other words, alcohol free. As you
15 can see from this table, it appeared that the
16 medium dose had borderline evidence of fewer days
17 of acamprosate, of complete abstinence based on
18 either the mean days or the median days.
19 Using the median as an example, you have
20 56 days for the medium dose compared to 78 days for
21 placebo having cumulative abstinence duration.
22 Similarly, for the percent of that, 38, much lower
23 compared to placebo. I will refer to these numbers
24 later.
25 Taken together, the total evidence based
152
1 on the five efficacy endpoints, there was no
2 evidence of medium-dose acamprosate effect on any
3 of the endpoints nominally although the high-dose
4 acamprosate appeared to perform better numerically
5 in the time to first heavy-drinking days.
6 [Slide.]
7 Thus, based on the prespecified primary
8 efficacy outcome, the result indicated that there
9 was no statistical evidence of this medium-dose
10 acamprosate. There were exploratory or supportive
11 analyses prespecified in the protocol. We
12 performed these analyses and could not find an
13 acamprosate medium dose effect.
14 [Slide.]
15 Right before the NDA, new drug
16 application, submission, the sponsor met with the
17 agency and acknowledged that the medium-dose
18 acamprosate failed to show a statistically
19 significant effect and submitted a new statistical
20 analysis plan. The highlight of this new plan
21 included the definition of the CAD was modified
22 post hoc. The algorithm of imputation on the
23 dropout patients was changed and the newly
24 considered outcome was percent abstinence duration.
25 Interestingly, the endpoint actually used
153
1 in the NDA submission was percent abstinence
2 duration but adjusted for treatment discontinuation
3 which appeared to be shorter in this medium-dose
4 acamprosate; that is, the variable, ALCCAD. This
5 endpoint was not included in the revised
6 statistical analysis plan although it was presented
7 at the pre-NDA meeting with the agency.
8 [Slide.]
9 What you have seen presented by the
10 sponsor is based on this Model No. 1. It contains
11 the seven covariates that Dr. Mason had explained.
12 I would like to just point you to the one
13 particular problematic variable, treatment
14 exposure. This model was discussed at the phase II
15 pre-NDA meeting but this model was not part of the
16 revised statistical analysis plan submitted at that
17 time.
18 [Slide.]
19 The sponsor was asked to also analyze the
20 data without that treatment exposure for Model No.
21 1, we just saw. The sponsor labeled it as Model
22 No. 2. Let's call it the six-covariate model.
23 Here, the treatment exposure was
24 calculated by multiplying the treatment compliance
25 and the treatment duration and then normalizing
154
1 into percent. It is worthwhile to note that the
2 treatment exposure so defined is potentially
3 treatment-related because that medium dose had a
4 higher percent of dropout rate and a shorter time
5 to discontinuation compared to the other two
6 groups.
7 In addition, such defined treatment
8 exposure variable is different from the baseline
9 variable and is not affected by the treatment
10 administration and the treatment outcome. But the
11 treatment exposure defined here would heavily
12 depend on when the treatment administration is
13 ended and whether patients comply with the
14 treatment assigned and why patients discontinue the
15 study.
16 [Slide.]
17 The CCAD outcome was the endpoint
18 discussed at the pre-NDA meeting. It was
19 prespecified but post defined. Of the two models
20 presented here, Model No. 1 and Model No. 2, using
21 the CCAD modified outcome, there were no
22 statistically significant findings of medium-dose
23 acamprosate.
24 Even if you don't do any adjustment, you
25 don't find anything either. Let us see how these
155
1 results can be drastically changed using the
2 post-hoc-defined primary-efficacy endpoint, ALCCAD.
3 Again, percent abstinence duration but adjusted for
4 treatment discontinuation.
5 [Slide.]
6 Here are the results using the
7 post-hoc-defined statistical model, the No. 1 and
8 No. 2 row, versus this model without further
9 covariate adjustment, the other four rows. Let's
10 look at the row labeled as mean No. 1 which was
11 based on seven covariates including the treatment
12 exposure, the problematic variable.
13 A nominal borderline statistical
14 significance was observed for the medium-dose
15 acamprosate compared to placebo, a p-value of
16 0.044. But when excluding that treatment exposure,
17 which is Model No. 2, such an acamprosate effect
18 disappeared, a p-value of 0.296. In contrast,
19 without this covariate adjustment, the unadjusted
20 mean showed a numerical trend of increased percent
21 abstinence duration from placebo to medium dose to
22 high dose, the third row here.
23 That is an adjusted mean. You can also
24 see on an adjusted median, the percent is
25 essentially the same between the medium dose and
156
1 the placebo of 59 percent.
2 I would like to bring to your attention
3 and clarify what the sponsor called an adjusted
4 mean or an adjusted median really is. As I just
5 mentioned, both the mean and the median was
6 adjusted for treatment discontinuation. In other
7 words, it rests strongly on treatment
8 discontinuation. Particularly, it was differential
9 among the three arms.
10 The truly unadjusted outcome was the CCAD,
11 the last two rows. As you can see, both the raw
12 mean and the raw median for acamprosate medium dose
13 was worse compared to placebo.
14 Let's put the high-dose acamprosate.
15 There was a numerically higher percent of
16 abstinence duration after adjustment for treatment
17 discontinuation. The high-dose effect appeared to
18 be shown nominally with the six covariate model and
19 was evident using the seven covariate model. These
20 better results did not hold up when we use the CCAD
21 outcome for the modeling.
22 [Slide.]
23 The sponsor considered four patient
24 populations to demonstrate the post hoc model. So
25 chosen, they were very consistent across the
157
1 patient population defined. As you have heard,
2 these are the four different patient populations;
3 the ITT, evaluable, multivariate ITT and
4 multivariate evaluable.
5 By showing this table, the nominal p-value
6 based on the seven covariate No. 1, all showed
7 statistical significance ranging from 0.044
8 borderline evidence to 0.008 significant evidence.
9 However, such evidence could not be supported when
10 the six covariate model No. 2 was applied to all
11 the four patient populations. None of them showed
12 statistical significance.
13 If a post hoc model is to be chosen
14 between Model No. 1 and Model No. 2, a less biased
15 analysis or a more persuasive analysis will
16 consider Model No. 2 without the treatment-exposure
17 variable. In addition, if these covariates are
18 really prognostic, including a fewer number of
19 covariates should still demonstrate some kind of
20 acamprosate medium-dose effect and should be
21 consistently reported in the literature cited by
22 the sponsor. But it did not.
23 [Slide.]
24 One might wonder what was the rationale
25 for the Model No. 1 chosen by the sponsor which was
158
1 not provided a priori. As previously shown, it was
2 the model with seven covariates that demonstrated
3 an acamprosate medium-dose effect but not the other
4 which excluded treatment exposure.
5 [Slide.]
6 This, of course, makes our job tougher.
7 We performed a few exploratory analyses. The idea
8 here was to understand how robust the results were
9 based on Model No. 1 chosen by the sponsor in the
10 NDA submission but not in the original protocol
11 analysis plan.
12 The exploration went on to include models
13 that always have the center in there or having one
14 variable at a time, or some combination of those.
15 This consisted of more than 30 models that we
16 tried. Other than the one model that the sponsor
17 identified, we found that there was no
18 statistically significant acamprosate medium-dose
19 effect from these various reasonable explorations
20 but there was one that works, which is the one that
21 included the abstinence goal and the
22 treatment-exposure variable together but not
23 individually.
24 [Slide.]
25 I would like to show you that, of the
159
1 seven covariates chosen by the sponsor, two of them
2 indicated potential imbalance between the three
3 treatment arms, namely treatment exposure and
4 abstinence goal.
5 As shown in this table, median exposure
6 was shorter in acamprosate medium-dose group
7 compared to the other two. This was consistent
8 with the shorter time to treatment discontinuation,
9 15 versus 20 or 21. In addition, there was a trend
10 in patient's baseline abstinence goal for the
11 treatments received as mentioned by the sponsor.
12 It appeared that numerically,
13 placebo-treated patients was more desirable to be
14 complete abstinence than acamprosate-treated
15 patients, 45, 40, 32. In contrast, if one
16 considered a more realistic goal of a slip is
17 possible versus others, the reverse numerical trend
18 was observed, 28, 31 to 39. It is the reverse
19 trend of the complete abstinence goal.
20 As pointed out by Dr. Winchell, when one
21 does not distinguish between complete abstinence
22 goal and the goal of allowed a slip is possible,
23 then there was essentially no imbalance among the
24 three treatment arms, as you can see, 73 percent,
25 71 percent.
160
1 [Slide.]
2 Here is a different way to look at the
3 data. In the following two figures, I will be
4 using green color to represent the medium dose,
5 darker blue for placebo and coral color for high
6 dose. For heavy drinking days, when the data was
7 summarized at each visit alone on the observed
8 data, as shown in this figure, it appeared that
9 acamprosate medium-dose group, the green color on
10 the top, showed a consistently larger number of
11 mean heavy-drinking days as compared to placebo.
12 Although the high dose had only one-third
13 of the patient size compared to the other two, an
14 apparent fewer number of heavy drinking days across
15 all the visits appeared to be evident and the
16 separation of the curve was consistent from Week 8
17 to Week 24, the end of the trial.
18 [Slide.]
19 In contrast, the distribution of any
20 drinking days at each visit was comparable among
21 the three treatment arms.
22 [Slide.]
23 From these various results shown, can we
24 conclude that the medium-dose acamprosate is
25 effective? First of all, the U.S. trial was
161
1 sufficiently powered to study the efficacy of this
2 dose but, clearly, there was no evidence of
3 medium-dose acamprosate when only one covariate was
4 accounted for. Even suppose that one covariate is
5 the potential outcome-related treatment exposure
6 alone. It didn't reach any statistical
7 significance.
8 To appropriately account for the
9 covariates, that should be unrelated to treatment
10 or outcome; that is, when that treatment-exposure
11 covariate is excluded from the model, we have shown
12 from a few example models, out of a total possible
13 128 models, the medium-dose acamprosate effect was
14 not found.
15 In addition, a numerically higher number
16 of heavy-drinking days relative to placebo at each
17 visit was observed.
18 [Slide.]
19 In fact, the 10 percent medium-dose effect
20 was highly dependent on post hoc selection of
21 covariates that were included in the model; for
22 example, a model including just two covariates, the
23 abstinence goal and the treatment exposure, or that
24 one model having all the seven covariates
25 coexisting in that model.
162
1 We have pointed out the problem with
2 models including the treatment-exposure covariate
3 because it could not be obtained until after
4 randomization of treatment assignment, after
5 treatment compliance and after treatment
6 discontinuation. An even more serious concern in
7 this exercise is the potential multiplicity
8 problem. In other words, could it be that the
9 sponsor performed analysis using only this
10 post-hoc-defined seven covariates or using many
11 more models to pick up this specific Model No. 1;
12 namely, what is the chance that one is going to
13 find a statistical significance after analyzing the
14 data using so many different models.
15 We all know that if one tests the same
16 parameter 100 times, five times are going to show
17 statistical significance simply based on chance
18 alone. Here, we found two out of 128.
19 [Slide.]
20 In this U.S. trial, the study was not
21 sufficiently powered to study this high-dose
22 effect. Rather, this dose was included to explore
23 the efficacy and safety. In a previous slide
24 showing mean heavy-drinking days, you have noticed
25 a numerically superior effect of acamprosate high
163
1 dose relative to placebo was seen at the later
2 visit of the treatment period and was consistent
3 throughout the end of the trial.
4 In addition, this high-dose effect
5 appeared to be seen if the adjustments always
6 included the abstinence goal but not otherwise. If
7 a model was performed using ALCCAD but not the
8 CCAD, we could not tell whether such finding was
9 real or by chance alone since the sample size was
10 only one-third of those powered for studying an
11 acamprosate effect.
12 [Slide.]
13 Here I would like to summarize the U.S.
14 experience. The medium-dose acamprosate appeared
15 to have worse dropout characteristics. The effect
16 of this medium dose was not shown based on the
17 protocol-specified primary efficacy outcome
18 although post-hoc-defined primary efficacy endpoint
19 of CCAD.
20 For the acamprosate medium dose, the
21 sponsor's post hoc chosen Model No. 1 or, for that
22 matter, Model No. 2, can be problematic as
23 statistical significance must rely on which
24 particular post hoc baseline defined covariates
25 and/or post randomization defined variables were
164
1 included in the model. The finding was very
2 fragile because the carefully chosen model showing
3 statistical significance could not hold its
4 significance after multiplicity adjustments.
5 [Slide.]
6 As for the high-dose acamprosate, the
7 exploratory analysis is suggested in the effect in
8 the time to first heavy drinking days and in the
9 mean heavy drinking days at each study visit over
10 the treatment period. Such heavy drinking days do
11 not adjust for treatment discontinuation like
12 ALCCAD.
13 It is emphasized, however, that the
14 finding in the high-dose acamprosate is simply
15 hypothesis generation as it didn't have sufficient
16 sample size for the study and had lack of safety
17 information for the dose level. The small sample
18 size prevented us from better understanding this
19 high-dose acamprosate treatment effect.
20 [Slide.]
21 So what is the difference between the
22 European and U.S. trials in terms of efficacy
23 outcomes? Why are we getting conflicting evidence
24 given randomizations were properly done. From the
25 statistical perspective, the biggest problem, in my
165
1 view, is the issue of differential dropout from the
2 study in terms of the time to discontinuation, in
3 terms of percent of dropouts and also in terms of
4 the distribution of reasons of dropouts.
5 We immediately face the problem of
6 differential dropouts in the opposite direction.
7 In other words, what have we found on the proposed
8 to-be-marketed acamprosate 2-grams-per-day effect?
9 We saw in the European trials, patient treatment
10 with acamprosate tended to stay in the trial longer
11 and less dropouts, but it was reversed in the U.S.
12 trial.
13 The sponsor had defined how they would
14 handle the missing data or data needed for the
15 dropout patients a priori but realized that it
16 didn't work and modified the definition after the
17 data had been collected when meeting with the
18 agency at the pre-NDA meeting and then modified
19 this outcome again as ALCCAD further by adjusting
20 for patient discontinuation.
21 Further data dredging was to include
22 treatment compliance and treatment duration to
23 create a variable called treatment exposure. That
24 can only be collected after the treatment
25 randomization. We believe that it is important and
166
1 there is a need to have a well-thought prespecified
2 algorithm for handling dropout patterns rather than
3 post hoc defined and redefined.
4 This concludes my review experience.
5 Thank you.
6 Questions from the Committee
7 DR. OREN: It is now time for the
8 committee to ask questions of the FDA regarding the
9 previous two presentations. Does anybody wish to
10 begin?
11 Dr. Rudorfer?
12 DR. RUDORFER: A question for Dr.
13 Winchell. We heard that about 10 percent of the
14 U.S. patient sample had undergone medical detox
15 before enrollment. Did you look at that subgroup
16 specifically?
17 DR. WINCHELL: I didn't because there were
18 so few of them. But I think that that was one of
19 Lipha's prespecified analyses so they may be able
20 to address that.
21 DR. GOODMAN: The statisticians can
22 correct me if I'm wrong, but I don't believe that
23 we had a prespecified plan for looking at the detox
24 patients. What we plan to do, patients were
25 stratified according to whether or not they had
167
1 undergone detox before they were randomized. But,
2 again, this was a surprising finding to us. We
3 expected that at least a third of the patients, if
4 not more, would undergo detox but, in fact, it was
5 only, as you saw, about 10 percent of patients.
6 DR. OREN: Dr. Hughes.
7 DR. HUGHES: Does anybody know, of all the
8 patients who come in for alcohol treatment, how
9 many of them are already abstinent at the time they
10 come in? Is there any kind of health-resources
11 database on that? Celia, do you know of any or do
12 the Lipha people know? Is that 90 percent of the
13 patients or 20 percent?
14 DR. WINCHELL: The best data I have ever
15 seen on that question was from Dr. Mason who
16 presented some very interesting data, I think from
17 this study, showing that people are really bad off
18 until they make the call to enter treatment and
19 then, between making the call and actually entering
20 treatment, they seem to do a little better.
21 But I think we have got lots of experts
22 here from NIAAA and Dr. Mason who may know
23 something about that.
24 DR. OREN: Dr. O'Brien?
25 DR. O'BRIEN: I think Dr. Winchell alluded
168
1 to the fact about the current American healthcare
2 system. In fact, it has really changed. We began
3 studying discontinuation in the 1970s and, at that
4 time, there were a lot of inpatient alcohol
5 detoxification programs and we actually did random
6 assignment between inpatient and outpatient in a
7 randomized clinical trial.
8 Nowadays, it is very difficult for us to
9 study this because it is so expensive. We have to
10 get an NIH grant to pay for the inpatient days
11 because there aren't any available through any
12 other system. So I think that things have really
13 changed and the modal method now is for alcoholics,
14 in the United States, at least, to come to us with
15 blood-alcohol levels fairly significant, sometimes
16 incredibly high because they are so tolerant and
17 they just walk in or drive up despite huge alcohol
18 levels.
19 Then we have to figure out how to get them
20 detoxed. Depending on what the protocol is, we may
21 have to find an inpatient program which is, as I
22 said, difficult or we do an outpatient detox.
23 DR. OREN: Any further questions from the
24 committee to the FDA? Dr. Schatzberg?
25 DR. SCHATZBERG: I have a question for Dr.
169
1 Winchell and Dr. Wang. It seems that, on your
2 reanalysis, that the European data are pretty
3 convincing in terms of what you would agree would
4 be a reasonable criterion for efficacy, I gather
5 from what you concluded.
6 But just as something for the committee or
7 for my edification, these studies were done a long
8 time ago, obviously. How do you feel about, in a
9 way, changing what is the specified outcome
10 criterion in a post hoc analysis in that way. In a
11 sense, are we doing something contradictory? We
12 are sort of, on the one hand, saying, in the U.S.,
13 we are going to throw out the EFF data because it
14 is post hoc, and whatever.
15 There are issues, there, granted. Yet we
16 are still sort of doing that except it is our own,
17 or the FDA's, reanalysis. What kind of criteria
18 would you use or would you recommend for what
19 should constitute a reanalysis and is part of it
20 just that these are so old in terms of the studies?
21 DR. WINCHELL: I will start and then I
22 will let Dr. Wang respond. First of all, the
23 difference between an efficacy evaluable post hoc
24 analysis and some of the other types of subset
25 analyses we did, as I mentioned, it has to do with
170
1 whether the subsets can be defined by
2 prerandomization variables.
3 The real problem with post hoc analysis,
4 the reason people tend to dismiss it, it that there
5 is the risk of multiplicity, the risk that, simply
6 by chance, if you do enough of them, you will get
7 one coming out statistically significant, as you
8 know.
9 Nevertheless, we do these types of
10 analyses to see whether there is differential
11 effect in women and men, differential effect by age
12 or by race. Usually, the studies are not powered
13 to generate a statistically significant difference
14 in any type of subset. They are powered just big
15 enough to demonstrate and effect in the ITT
16 population.
17 So we don't expect these analyses to come
18 out with a statistically significant result. We
19 expect them to give us some trends or some
20 understanding or just to shed some light on who in
21 the population is particularly prone to benefit of
22 not to benefit.
23 We do these routinely. Rarely one might
24 take as the body of evidence supporting an
25 application some type of post hoc reanalysis of
171
1 data as supportive. If you had one or two very
2 strong studies, you might look retrospectively at
3 existing datasets in a way that was not anticipated
4 at the time the data was collected and say that
5 this analysis generates supportive, confirmatory
6 evidence that helps to complement the other results
7 and complete the body of evidence necessary for
8 regulatory decision making.
9 So it is not uncommon to look
10 retrospectively at older sets of data. Usually, we
11 get a little uncomfortable if that is the only
12 basis on which the efficacy can be concluded. I
13 think of this, and I know Dr. Wang maybe thinks of
14 this differently because she is a statistician and
15 I am a medical officer, but I think, in some ways,
16 of approaching this European data the way one might
17 approach a literature-based application where there
18 is this large body of data. I have got the actual
19 data. I can look at it various ways.
20 I think what we hoped to get when we first
21 met with Lipha was what I described, that we would
22 have one American trial that was successful but
23 that could not stand alone--it was not
24 replicated--and that we would accept as
25 confirmatory evidence analysis of older European
172
1 data notwithstanding the fact that it was a
2 different dose and a different dosage regimen and
3 that those pieces together would form the basis of
4 our decision.
5 Ultimately, we were faced with going
6 forward without that successful American study and
7 we still tried to make what we could out of the
8 European data.
9 I don't know if that addresses your
10 question. I will also ask Dr. Wang to talk about
11 how she sees it statistically and I see that my
12 boss wants to tell you what she thinks of it. So I
13 will let her go first.
14 DR. McCORMICK: Thank you. I guess,
15 really, the crux of your question is how is it that
16 we can go into the U.S. dataset and do these post
17 hoc analyses ourselves and not accept what the
18 sponsor has given us in terms of their post hoc
19 analyses, and yet we are taking the European
20 dataset and saying we are all going to do a post
21 hoc analysis here, and that is going to be the
22 basis of our regulatory decision.
23 Yes, that does give us some discomfort.
24 Let me first say that, as far as the U.S. post hoc
25 analyses are concerned, I think both on the part of
173
1 the sponsor and ourselves, is that these are purely
2 hypothesis-generating. We are looking to try to
3 understand this information, not to draw any
4 conclusions about it, because we feel quite
5 comfortable that we cannot use the United States
6 study in making a regulatory decision.
7 That leaves us with the bulk of the data
8 from Europe, or all of the data from Europe, to
9 make our decision about. Yes; it does give us some
10 discomfort in seeing trials in cases where we
11 haven't had prespecified primary-outcome measures
12 and we have to reconstruct them based upon what the
13 trial objectives were.
14 Yet, when we take the most conservative
15 approach, even more conservative than what was
16 probably originally intended, looking at complete
17 abstinence, it is consistent across all the
18 studies.
19 This, truly, is something that we would
20 like to bring to the table, though. But I think
21 even beyond having done that and taking the more
22 conservative approach, looking at complete
23 abstinence as an outcome, our even greater level of
24 discomfort and, really, the reason for having this
25 meeting is not so much have we chosen a post hoc
174
1 analysis to do on this dataset but what is the
2 credibility of the dataset, itself.
3 Can we rely upon, for example, a one-year
4 study in which there have been only six visits,
5 where the data is largely imputed? Can we believe
6 that and can we base our regulatory decision on
7 these studies? That is the crux of the matter.
8 DR. WANG: I am going to talk about from
9 the statistical perspective. In terms of the
10 timing of the European trials versus the U.S.
11 trial, yes, we are going to say these are all post
12 hoc analyses. What you see from the European
13 studies, you have all the consistencies across all
14 the outcomes that you looked at.
15 When there is a problem of differential
16 dropout between the acamprosate and the placebo, it
17 is in the direction, you believe the drug works.
18 However, in the U.S. trial, the troubling thing is
19 the post hoc nature of it.
20 First of all, if the drug works, if the
21 prespecified analysis works, we don't need to talk
22 about the post hoc. So, going to post hoc, you
23 already failed the first step. In that post hoc
24 situation, yes, we accept some kind of post hoc
25 evaluation. But, you start with one covariate
175
1 adjustment. It was believed by the sponsor that
2 the abstinence goal was a very prognostic one. If
3 you have a model, just include treatment center and
4 that covariate of complete abstinence goal, you
5 don't find the statistical evidence.
6 If you then say, all right, let me look at
7 treatment center and the slip is okay, because that
8 is also differential in the opposite direction,
9 still you did not see the statistical evidence.
10 Even if you adjust for just one covariate,
11 treatment exposure, it is not there either.
12 So this post hoc nature was trying to
13 explain what is going on. You would expect that if
14 the effect is really there, then, using a fewer
15 number of covariates should still give you some
16 kind of treatment-effect size. But it wasn't in
17 this case.
18 So the post hoc nature, in this particular
19 situation, is very troubling.
20 DR. OREN: Dr. Schatzberg?
21 DR. SCHATZBERG: I appreciate the answer.
22 It was really more kind of a structural--as Dr.
23 McCormick raised. But let me ask one other
24 structural one, if I might, because of something
25 that Robert raised before, and that is, while this
176
1 is somewhat of a different division, I guess, of
2 the FDA from the psychopharm group, is there
3 concern in the agency that recommending approval
4 based on the European portfolio and without U.S.
5 data would not jive or go with other efforts on the
6 part of this committee.
7 I am not a member of the committee so I
8 just raise that as a precedent, or is that just
9 because they are different illnesses and different
10 agencies and different criteria?
11 DR. McCORMICK: To answer your question,
12 there really are no concerns on the part of the
13 agency about making a regulatory decision based on
14 purely European data. As long as they are rigorous
15 and credible and the studies have been done using
16 good clinical practices and they are in sites where
17 we can do inspections.
18 In this case, there are. There have been
19 precedents where European data has been relied
20 upon. That is not an issue.
21 DR. OREN: Dr. Leon?
22 DR. LEON: Are there standards that the
23 agency uses for maximum dropout rate in clinical
24 trials? I mean, these dropout rates typically were
25 never less than 35 percent but, typically, 50 or 60
177
1 percent dropout.
2 DR. WINCHELL: These are not unusual
3 dropout rates for addiction-treatment trials. If
4 we had standards for unacceptable dropout rates, I
5 don't think we would be able to do
6 addiction-treatment trials that lasted more than
7 two weeks.
8 DR. OREN: Dr. Winchell, I wonder if you
9 could just say a little more, just specifically
10 focussing on the European studies and not focussing
11 right now on the broader question of approval but
12 just specifically on the efficacy of acamprosate in
13 the European studies? How would you summarize your
14 analysis?
15 DR. WINCHELL: Well, let me say that,
16 based on the data that I had available to analyze,
17 the very short three-month Pelc study certainly
18 showed an effect of acamprosate on complete
19 abstinence. The Paille study was more marginal on
20 that and the PRAMA study showed and effect if you
21 imputed failure to all the dropouts and not
22 necessarily if you didn't.
23 So, on complete abstinence, it looks
24 promising but it is not a blockbuster. In my own
25 made-up, what else can I do besides cumulative
178
1 abstinence duration analysis, there is a difference
2 between the dose proposed for marketing and placebo
3 in favor of acamprosate in all the studies. As I
4 mentioned, that is again driven primarily by the
5 completely abstinent subjects who, in this
6 analysis, have failure imputed after dropout.
7 So I can certainly get a good result.
8 But, obviously, I have some reservations about how
9 much I should believe my own analysis. I don't
10 mean to sound disrespectful about these studies.
11 All I know is that the American study reported 100
12 volumes and some of the European study reports are
13 one volume.
14 So I just have so much more detail
15 available for my scrutiny for the American study.
16 That is what we are accustomed to, actually, is
17 something on the order of the 100 volumes per
18 study. I should say that the case-report forms
19 were submitted electronically as were the
20 case-report tabulations so those weren't even
21 included in there.
22 That is the type of thing we are
23 accustomed to having available for our examination
24 and we didn't have that for the European data. So
25 that is why we are here.
179
1 DR. OREN: Dr. Cook
2 DR. COOK: I think you have covered this,
3 but just to clarify for me. If you took the
4 predefined outcome variable and the predefined
5 analysis by the sponsor, number one, were those
6 defined? Is there any doubt about whether they
7 were defined? In other words, do you have a
8 document that clearly specifies it. And, for those
9 three trials, what happened with those primary
10 hypotheses and their primary analyses?
11 DR. WANG: Are you specifically talking
12 about just the European studies?
13 DR. COOK: Yes; just the European studies.
14 DR. WANG: As Dr. Winchell mentioned, the
15 European-study information given to us was limited.
16 So that is why, in my presentation, I only based it
17 on percent complete abstinence and not others. So
18 I cannot make too much out of what I have--I mean,
19 in addition to what I have
20 DR. COOK: Okay. But, by limited, do you
21 mean that, in each trial, you couldn't see in their
22 documents that they had written a document before
23 the study started about what the predefined
24 analysis would be and what the predefined outcomes
25 were and did you have the data to see whether those
180
1 trials were positive given that standard.
2 DR. WANG: I think, from our internal
3 discussion while we were doing this priority
4 review, we had a discussion as to how much can we
5 believe in the European data in terms of the number
6 of days that the patients were abstinent.
7 As Dr. Winchell presented, there were--if
8 you are talking about Pelc II, it is biweekly
9 visits. But, for others, is a one to three months
10 kind of difference. So if you are doing
11 imputation, there is big chunk of time that you can
12 impute by days. Therefore, it was believed that
13 the quality of the data with those were
14 questionable and that was the reason of the focus.
15 DR. WINCHELL: I have something else I can
16 say to address your question. At least one of the
17 studies--I am thinking it is Pelc II--it said that
18 the primary outcome variable, the main criterion of
19 judgment, would be abstinence. But what it didn't
20 have in the protocol was any operationalization of
21 how that would be evaluated.
22 As you have seen, if your main criterion
23 of judgment is abstinence, you could look at time
24 to first drink, time to first heavy drink, time to
25 relapse, cumulative abstinence duration or any
181
1 number of other measures of abstinence. So then,
2 appended to the protocol, we then had a statistical
3 report. In the statistical report, it was set
4 forth what analyses were done. At least one of
5 them was a blinded analysis. I can say that much.
6 So one could assume that the statistician
7 decided what to do first. It is unclear. But it
8 is not like what we are accustomed to seeing in an
9 American NDA in 2002.
10 DR. WANG: I would like to add to that is
11 the difficulty in analyzing the Paille study. In
12 fact, the patient's dropout reasons were
13 reclassified even though those data were used to
14 have a European approval. By using the new defined
15 reasons of dropout and looking at the three
16 treatment-arm comparisons, you can get a different
17 result.
18 DR. OREN: Dr. Mehta?
19 DR. MEHTA: One way to look at it would be
20 that there are very few areas of medicine where you
21 do fourteen placebo-controlled studies and you turn
22 out to be a winner fourteen times. What the
23 sponsor has done is, in European, twelve or
24 thirteen times, rolled the dice against placebo and
25 it came out as a winner.
182
1 By the law of averages, I would have
2 expected the next trial will be negative and what
3 they did is essentially ably demonstrated the law
4 of averages works.
5 DR. OREN: Dr. McCormick?
6 DR. McCORMICK: I would just like to point
7 out that we were only given full study reports of
8 three of the European studies. We know that they
9 haven't all succeeded and I don't believe that they
10 all had complete abstinence as an outcome.
11 DR. OREN: I think we will take Dr.
12 Rudorfer with the last question and then we will
13 take our lunch break.
14 DR. RUDORFER: I am sorry to have to
15 compete with lunch. Just a couple of questions.
16 We have all been talking about the fact that
17 European studies are a decade old. I am wondering
18 if we have learned anything in the interim. For
19 instance, are the postmarketing data available that
20 might be informative just in terms of do people
21 actually refill their prescriptions over a year's
22 duration, issues like that?
23 DR. WINCHELL: Obviously, Lipha has much
24 more information than we do, but I just know
25 recently looking at some of their materials, that
183
1 it said market research shows that typical duration
2 of use was, like, three to six months. So it
3 doesn't sound like people are typically using it
4 for a year or more. But, certainly, I will let--I
5 see heads shaking but I did read that in the NDA
6 yesterday.
7 DR. CHABAC: I just want to remind you
8 that alcohol-dependent patients are very badly
9 compliant patients. To keep them treated for six
10 months with the treatment, I think it is a very
11 good sign that this drug could be beneficial to
12 them.
13 I told you that we have 1.5 million
14 patient years experience with the product. That
15 means that there are a lot of patients treated with
16 acamprosate. We have the experience with the NEED
17 Program where we treated nearly 2000 patients in
18 Europe. Dr. Mann can tell me if I am wrong, but I
19 think there is a benefit using that drug. It is
20 not a magic product but I just want to remind you
21 that the two drugs available on the market to treat
22 these kinds of patients have neither a very huge
23 rate of efficacy and that if we can bring something
24 safe to treat, to help, those patients, this is
25 something.
184
1 DR. MANN: I think to understand these
2 figures, the recommendation in Germany, at least,
3 is to give it for six months. All the doctors know
4 it would be given for six months and not for a year
5 or more which is now something that is recommended.
6 So if you have figures that show that it
7 is taken five or six months, this shows compliance
8 of the doctors, if you want.
9 DR. OREN: Dr. McCormick?
10 DR. McCORMICK: Just a word of caution
11 that I would like to insert and that is while it
12 may be important to understand how a drug plays out
13 in the postmarketing period, we would not accept a
14 postmarketing uncontrolled experience as evidence
15 of a product's efficacy as part of our making of a
16 regulatory decision.
17 DR. OREN: Before we break for lunch, I am
18 reminded to remind each member of the committee
19 that, because this is a public hearing, over the
20 one-hour lunch break, we are not supposed to talk
21 about any of this particular material because it is
22 out of the public forum. There will be plenty of
23 time later this afternoon to continue and we will
24 be back in one hour.
25 Thank you.
185
1 [Whereupon, at 12:15 p.m., the proceedings
2 were recessed to be resumed at 1:15 p.m.]
186
1 A F T E R N O O N P R O C E E D I N G S
2 [1:30 p.m.]
3 Open Public Hearing
4 DR. OREN: We are now ready to begin the
5 Open Public Hearing on today's agenda. The first
6 speaker is Dr. Victor Hesselbrock, Vice President
7 of the Research Society on Alcohol.
8 Dr. Hesselbrock?
9 DR. HESSELBROCK: Good afternoon. I am
10 Victor Hesselbrock, Vice President of the Research
11 Society of Alcoholism. I am also a Professor in
12 the Department of Psychiatry, University of
13 Connecticut School of Medicine and I am Director of
14 the Alcohol Research Center at the University of
15 Connecticut.
16 At this time, I have no financial interest
17 in Lipha Pharmaceuticals or any pharmaceutical
18 company but, as Director of the Alcohol Center, I
19 will indicate to you that two individuals, Dr.
20 Stephanie O'Malley and Dr. Henry Kransler, have
21 conducted studies of both naltrexone and
22 acamprosate and have received some remuneration
23 from the pharmaceutical companies. But they are
24 indirectly related to me. I am the Executive
25 Director and I am not associated with those
187
1 studies.
2 The Research Society on Alcoholism
3 appreciates the opportunity to present its views
4 about the importance of finding effective
5 pharmacological treatments for individuals
6 suffering from the psychological, social and
7 biomedical consequences of abusive drinking.
8 The RSA is a professional scientific
9 society of over 1400 members who are committed to
10 understanding and intervening in the negative
11 consequences of alcohol abuse through basic
12 research, clinical protocols, psychosocial research
13 and epidemiological studies. About one-third of
14 RSA members are also clinicians actively involved
15 in the treatment of individuals with
16 alcohol-related problems.
17 As we heard this morning, the cost of
18 alcohol abuse and dependence on American society
19 and individual lives is staggering. The cost to
20 the nation is estimated at approximately $185
21 billing annually. Not only are the fiscal costs
22 real and powerful, but alcohol misuse is costly in
23 many ways.
24 Estimates of alcohol-use disorders ranging
25 from abuse through dependence from the National
188
1 Longitudinal Alcohol Epidemiological Survey
2 indicates that about 7.5 percent or 14 million
3 Americans are affected. Further, a Robert Wood
4 Johnson Foundation report indicates that more than
5 700,000 people receive alcoholism treatment on any
6 given day. Approximately only 15 percent receive
7 inpatient treatment and these patients often have
8 the most severe form of alcohol problems.
9 The remaining patients receive outpatient
10 treatment from a variety of different treatment
11 providers including psychiatrists, primary-care
12 providers, psychologists, social workers and
13 self-help groups such as Alcoholics Anonymous.
14 Based on Project MATCH data, approximately 40 to 50
15 percent of those in outpatient treatment are able
16 to abstain in the first week of therapy but many
17 relapse shortly thereafter.
18 Although the combination of behavioral
19 therapies and currently available medications such
20 as disulfiram and naltrexone help 40 to 70 percent
21 of persons with alcoholism either reduce their
22 alcohol consumption or maintain abstinence up to
23 six months following treatment.
24 The relapse within one year of treatment
25 still ranges from 30 to 50 percent. The primary
189
1 reason for relapse to abusive drinking is
2 noncompliance with both the pharmacologic as well
3 as the behavioral treatment.
4 Importantly, and I think this is something
5 that has not been mentioned this morning to date is
6 a significant number of adolescents and young
7 adults are frequent consumers of large amounts of
8 beverage ethanol with disastrous consequences.
9 These are individuals that also would benefit from
10 new therapies.
11 A recently released report on college
12 drinking sponsored by the National Institute of
13 Alcohol Abuse and Alcoholism reveals that 1400
14 college students between the ages of 18 to 24 die
15 each year from unintended alcohol-related injuries.
16 An additional half a million students per year
17 between the ages of 18 to 24 are unintentionally
18 injured under the influence of alcohol. The
19 majority of these individual have not developed
20 physical dependence as discussed in some of the
21 studies this morning and typically do not seek
22 treatment. But, still, these are individuals that
23 would benefit from new therapies.
24 Alcohol abuse and alcohol dependence are
25 cites as major causes of medical morbidity, mental
190
1 retardation, accidental death and injury, homicide,
2 suicide, lost productivity and disruption of
3 family. Further, frequent and prolonged heavy
4 drinking contributes to illness in each of the top
5 three causes of death, heart disease, cancer and
6 stroke.
7 Chronic alcohol abuse is linked to nearly
8 half of all cirrhosis deaths, the tenth-leading
9 cause of death in the U.S. For some special
10 populations of American society such as Native
11 Americans and African Americans, the costs
12 associated with alcohol misuse are
13 disproportionately higher and may be directly
14 linked to some of the major health problems in this
15 group such as hypertension and diabetes.
16 The Indian Health Service estimates that
17 age-adjusted alcoholism mortality for American
18 Indians is 63 percent higher than the rate for all
19 other ethnic groups in the U.S. Overall, alcohol
20 mortality rates are particularly higher among
21 African-American men even though alcohol use tends
22 to be moderate for African Americans compared to
23 Caucasians and Hispanics.
24 Given the range and diversity of the
25 severity of alcohol problems across the general
191
1 population of the U.S., the number of available
2 medical treatments is extremely limited. As we
3 heard this morning, there were only two types of
4 medications and, in fact, only two medications that
5 are FDA approved, and that includes disulfiram
6 which is an aversive agent available since the
7 early 1950s and, more recently, naltrexone which is
8 the first medication approved by the FDA for
9 alcoholism treatment in nearly 50 years.
10 Compliance with both these medications is a problem
11 but, when combined with behavioral therapy, both
12 have been shown to be useful in reducing drinking
13 in selected but not all patient groups.
14 However, medication is not without its
15 limits in relation to safety of use. Neither
16 disulfiram nor naltrexone, for example, are
17 recommended for individuals with significant liver
18 injury or liver disease such as cirrhosis or
19 hepatitis C. Given that alcohol is a known
20 hepatotoxic agent, many individuals who desperately
21 need to quit drinking in order to improve their
22 health are not candidates for these medications.
23 Alternative treatments that are not
24 hepatotoxic and that can be safely used by
25 medically compromised patients are critically
192
1 needed. A larger number of medical treatments are
2 required given that no one pharmacological
3 treatment is strongly effective and probably helps
4 only a subgroup of patients.
5 Currently, members of the RSA and other
6 scientists are conducting both basic and clinical
7 trials on a number of promising compounds to
8 identify effective pharmaceutical agents to treat
9 individuals with alcohol dependence or those who
10 chronically abuse alcohol. The RSA asks that you
11 give careful consideration to the current proposal
12 for approval of acamprosate as the currently
13 available clinical armamentarium is quite sparse
14 and is really insufficient to address the very
15 needs of the treatment providers across the
16 spectrum of alcohol-related problems that they are
17 asked to treat.
18 Thank you for the opportunity to present
19 our views.
20 DR. OREN: Thank you.
21 Has Dr. Johnathan Chick arrived? No?
22 Then we will go on. The next Open Public Hearing
23 presenter is Dr. Steven Mirin, Medical Director of
24 the American Psychiatric Association.
25 DR. MIRIN: Thank you, Mr. Chairman,
193
1 members of the advisory committee. I am Steve
2 Mirin, Medical Director of the American Psychiatric
3 Association, a medical specialty society
4 representing more than 38,000 psychiatric
5 physicians nationwide.
6 I commend the FDA and this committee for
7 undertaking a review of the efficacy of acamprosate
8 for the treatment of alcohol dependence. I have no
9 association with any pharmaceutical company that
10 develops or distributes this drug.
11 I come before you not as an expert on the
12 pharmacology of acamprosate but as the
13 representative of 38,000 care-givers concerned
14 about the public-health need for more effective
15 treatment for alcoholism. Alcohol, as you know,
16 remains the commonly abused drug by youth and
17 adults alike in this country. About 14 million
18 Americans meet medical criteria for the diagnosis
19 of alcohol abuse or dependence and 40 percent of
20 Americans have direct family experience with the
21 illness.
22 The financial burden of alcohol abuse and
23 dependence is estimated at $185 billion a year, 52
24 percent greater than the estimated cost of all
25 illegal drug use and 21 percent greater than the
194
1 estimated cost of smoking-related problems. More
2 than 70 percent of this amount is attributable to
3 lost productivity and lost earnings, but the
4 medical costs are also staggering. Up to 40
5 percent of patients in urban hospital beds are
6 there for the treatment of conditions caused by or
7 exacerbated by alcohol including diseases of the
8 brain and liver, certain forms of cancer, accidents
9 and violence.
10 These data underscore the need for more
11 effective clinical interventions in people
12 suffering from alcoholism. In this context,
13 approval of the use of acamprosate, a drug shown in
14 numerous international studies to be effective in
15 the maintenance of abstinence and relapse
16 prevention in patients with a history of alcohol
17 dependence would be, in our view, in the interests
18 of this large patient population and an important
19 new tool for the practitioners I represent and for
20 other healthcare providers across the country
21 As you know, acamprosate is currently
22 approved for use in 39 countries and about 1.5
23 million persons with alcohol dependence have been
24 treated worldwide. The drug appears to be well
25 tolerated with no serious adverse side effects and
195
1 no evidence of abuse potential or rebound effects
2 when discontinued. It can be used safely in
3 patients with liver disease and it does not impair
4 performance on motor tasks like driving. It has a
5 very high margin of safety.
6 Multiple controlled clinical trials have
7 demonstrated the efficacy of acamprosate in
8 reducing craving for alcohol and helping maintain
9 abstinence in previously dependent patients. This
10 is not a trivial finding. It can reduce the time
11 to first drink. There is a higher rate of complete
12 abstinence, a greater percentage of abstinent days
13 while on medication and these effects are sustained
14 over post-treatment follow-up periods for as long
15 as one year in some studies.
16 There are fewer hospitalizations for
17 detoxification and diminished need for
18 rehabilitation in institutional settings and a
19 diminished rate of relapse to heavy drinking or
20 even sporadic drinking. As one of the
21 investigators in the early studies of naltrexone, I
22 can well appreciate the need to avoid slips in
23 alcoholics. Slips are not trivial events. They
24 are the forerunner of relapse.
25 Not surprisingly, a study conducted in 600
196
1 outpatients with alcohol dependence in this country
2 indicated that patients who were not motivated to
3 be abstinent are not as likely to benefit from
4 acamprosate whereas those who were significantly
5 more likely to meet their treatment goals when
6 compared to folks given placebo. This suggests
7 that, as in other addictive disorders,
8 psychotherapy is just one aspect of a successful
9 treatment program.
10 In summary, we believe that on the basis
11 of the findings to date, acamprosate has
12 demonstrated efficacy in the treatment of alcohol
13 dependence and has provided a cost-effective
14 treatment for these patients. Given the high
15 prevalence of alcoholism in this out and the
16 medical, economic and emotional costs of these
17 disorders, approval of acamprosate can have
18 important benefits for millions of our citizens and
19 for our society as a whole.
20 Thank you for the opportunity of
21 presenting this testimony.
22 DR. OREN: Thank you, Dr. Mirin.
23 Our next public speaker is Dr. Edward
24 Eder, Medical Director of the Comprehensive
25 Addiction Treatment Program, Fairfax, Virginia.
197
1 DR. EDER: Good afternoon, Mr. Chairman
2 and panel members of the advisory committee. I
3 appreciate the opportunity to speak on this
4 subject.
5 My name is Edward Eder. I am an internist
6 with twenty years practice predominantly in the
7 field of addiction medicine. I am a consultant to
8 Fairfax County's Alcohol and Drug Services, a
9 member of the American Society of Addiction
10 Medicine and Medical Director of the Comprehensive
11 Addiction Treatment Services.
12 As an internist and Medical Director of
13 the Comprehensive Addiction Treatment Services
14 affiliated with INOVA Fairfax Hospital, I have been
15 aware of the high risk of relapse in patients with
16 alcohol dependence despite involvement in
17 well-designed outpatient treatment or in sober
18 structured environments. With the advent of
19 greater understanding of the neurochemistry of the
20 addicted brain, I share the hope that
21 pharmacological agents would become available to
22 assist patients in maintaining abstinence.
23 Our current list of medications to reduce
24 relapse is very limited and acamprosate would be an
25 important addition to therapeutic options. There
198
1 are three specific categories of patients who would
2 most benefit from acamprosate in terms of our
3 clinical practice. One, patients on opioids who
4 are not candidates for naltrexone and would benefit
5 from an agent that would assist alcohol abstinence.
6 In methadone-maintenance programs, up to 50 percent
7 of patients have alcohol-dependence or alcohol-use
8 disorders for, instance.
9 Also, patients with hepatotoxicity
10 excluding Child Class C category who may not
11 qualify for disulfiram or naltrexone as well as
12 patients who might benefit from the neuroprotective
13 effect of acamprosate such as individuals with
14 alcohol-withdrawal seizures.
15 The addition of acamprosate to the
16 available medicines for treatment of alcohol
17 dependence would allow for future combinations that
18 may afford greater efficacy. Given the novel
19 pathways which acamprosate appears to act upon, the
20 potential for additive or, perhaps, synergistic
21 effects is promising.
22 I believe that there is a strong clinical
23 justification for a medication such as acamprosate
24 and believe multicenter trials in Europe appear to
25 confirm both efficacy and safety. I urge the panel
199
1 to consider the approval of the medication for the
2 treatment of addiction.
3 Thank you very much.
4 DR. OREN: Thank you, Dr. Eder.
5 Since this is an Open Public Hearing, I
6 wanted to ask if there is any member of the general
7 public here who wishes to make a statement in
8 regard to the topic at hand.
9 Please. Do you want to introduce
10 yourself?
11 DR. PUBLICKER: My name is Mark Publicker.
12 I was actually on the comment list. I am the Chief
13 of Addiction Medicine for Kaiser Permanente in the
14 MidAtlantic Region. I am also President of the
15 Virginia Society of Addiction Medicine.
16 I am speaking on behalf of the Chiefs of
17 Addiction of Addiction Medicine for Kaiser
18 Permanente nationally. We provide care to over 10
19 million Kaiser Permanente members coast-to-coast.
20 I am also speaking on behalf of Virginia's
21 addiction-medicine specialists. I am also speaking
22 on behalf of my alcoholic patients many of whom are
23 desperate for an effective medical treatment for
24 this disabling behavioral disorder.
25 Following the lead of earlier speakers, I
200
1 should hasten to add I have no financial interest
2 in Lipha and, quite frankly, I don't have any
3 investments that will help me pay for my daughter's
4 college education next year. So I am clean.
5 Since its FDA-approved indication for the
6 treatment of alcoholism, I and my local partners
7 have prescribed naltrexone to thousands of
8 alcoholic patients. I am proud to say I appear to
9 hold the record. We have found it to be very
10 effective in combination with behavioral therapies
11 and decreasing craving and relapse allowing our
12 patients to focus their energies on psychosocial
13 treatments rather than on white-knuckling their
14 recovery.
15 We have found that patients are grateful
16 for such psychotherapy much in the same way that
17 chronic heartburn sufferers are grateful the first
18 time they are prescribed proton-pump inhibitors. I
19 have received many phone calls of the same quality.
20 I would like to also add that I have a number of
21 patients who schedule follow-up visits with me
22 every six months for the last few years checking on
23 the status of acamprosate because they are getting
24 incomplete relief when they are on their
25 naltrexone.
201
1 Nonetheless, many patients cannot take
2 naltrexone. Some develop intolerable
3 gastrointestinal side effects that prevent its use.
4 Methadone-maintained patients and chronic-pain
5 patients on long-term opioid therapy with
6 co-occurring alcoholism cannot take naltrexone.
7 Finally, despite dosages of 100 to 200 milligrams
8 daily, some patients continue to experience both
9 craving and relapse.
10 I have carefully reviewed the European and
11 North American literature on acamprosate. There is
12 extensive documentation of its superiority to
13 placebo in promoting enhanced abstinence and early
14 recovery. Acamprosate has an excellent safety
15 profile and there is some suggestion it may have a
16 neuroprotective effect. Studies have shown
17 acamprosate and naltrexone, taken together, have an
18 additive effect in promoting abstinence.
19 I urge the panel to consider the millions
20 of lives that will benefit from the addition of
21 such an effective new treatment for such a
22 devastating disease and approve acamprosate.
23 Thank you.
24 DR. OREN: Thank you.
25 Any other general comment from the public?
202
1 Charge to the Committee
2 DR. OREN: I will now call upon Dr.
3 Cynthia McCormick to deliver the charge to the
4 committee.
5 DR. McCORMICK: Thank you, Dr. Oren. This
6 morning, you have heard from Lipha and from the FDA
7 on the four clinical trials in question. I would
8 like to remind you that this advisory committee
9 meeting today will not be one in which a final
10 approval recommendation is being requested.
11 Recall that there are other aspects of the
12 drug-approval decision which are not being brought
13 for discussion today. The drug safety, as I
14 mentioned earlier, is still under evaluation and is
15 expected to be completed by the end of this month.
16 Both clinical inspections and inspections of the
17 manufacturing sites have also not been done yet.
18 In fact, one of our inspectors is here today and
19 will be leaving for France this afternoon to begin
20 his inspection of some of the European sites.
21 So these will both have to be weighed into
22 the decision for approval and in the timing of
23 approval, potentially.
24 We are asking you to assist the FDA in
25 assessing the weight of the evidence provided in
203
1 support of the efficacy of this product. A number
2 of exploratory analyses have been performed in an
3 effort to understand or explain the discrepant U.S.
4 results both by the FDA and by Lipha. You should
5 regard these analyses not as definitive but as
6 hypothesis-generating.
7 The FDA, in the end, does not accept the
8 results as positive nor feel that they should be
9 weighed in the decision for approval nor does the
10 FDA have an explanation for the failure of the
11 trial. So where does that leave us? It leaves
12 with questions about whether the populations are so
13 different that the European results may not apply,
14 about whether the differences in methodology alone
15 account for the successes of the European studies
16 and, therefore, whether the effect was real.
17 The effectiveness standards for approval
18 of a new molecular entity include at least two
19 adequate and well-controlled studies that
20 demonstrate a significant effect on the outcomes
21 that have been determined to demonstrate a
22 clinically meaningful result regardless of the
23 trial's origins, European or U.S., of course with
24 the caveat, as I mentioned earlier today, that the
25 sites are those that can be inspected. So the fact
204
1 that the bulk of the experience, the efficacy
2 experience, is European is not a problem for the
3 FDA.
4 The standards require a certain level of
5 quality such as the existence of a prospective plan
6 to assure data quality, availability of source
7 documents that can be used to verify the quality of
8 the data and the accuracy of the data and conduct
9 of the study following the standards of good
10 clinical practice. As is the agency's practice,
11 there will be inspections, as I mentioned, to
12 evaluate the veracity of the data.
13 As alluded to earlier, there is the
14 question of the credibility of the approach of
15 using highly imputed data in the European studies.
16 This should be carefully considered when assessing
17 the value of these studies. We will ask you to
18 reflect on all that you have heard, consider the
19 totality of evidence giving consideration and
20 weight to such factors of quality of data, strength
21 of the effect size and, most importantly, whether
22 the results that are positive are credible.
23 At the end of the day, the FDA must have
24 confidence that its decision will be based on
25 information that cannot be questioned.
205
1 So, in returning to the meeting, we ask
2 you to deliberate on the following questions, and I
3 will read them to you. Given the conflicting
4 results between the European studies and the
5 American study, is there sufficient evidence of the
6 efficacy of acamprosate in the treatment of
7 alcoholism to warrant approval? In this, consider
8 not only the quantity but also the quality of the
9 evidence provided in support of the effectiveness
10 claim.
11 How can the discrepant results be
12 reconciled or do they need to be? Finally, do the
13 data support any conclusions regarding subgroups of
14 patients more likely to benefit from acamprosate?
15 Please discuss that.
16 Thank you very much.
17 Continuation of Discussion
18 DR. OREN: Before the committee begins its
19 open discussion, I have a few questions that I
20 wanted to ask of the sponsor to help eliminate some
21 of our discussions. Could you, perhaps, clarify
22 what was your NDA strategy for this drug?
23 DR. GOODMAN: As I mentioned earlier this
24 morning, our NDA strategy was always planning to
25 use the European dossier as a substantial part of
206
1 our database. We always intended to use at least
2 two of the European studies as fulfilling the
3 requirements that Dr. McCormick has just mentioned,
4 adequate and well-controlled, and, in addition, we
5 felt it incumbent upon us to also preform a study
6 in the United States to confirm both the efficacy
7 as well as get further information on safety in a
8 broader population.
9 When the U.S. study results for the ITT
10 population did not show a difference between
11 treatment and placebo, our strategy was redefined
12 in terms of the amount of European data that we
13 were going to use in that we decided to add an
14 additional study to what we considered to be our
15 pivotal study.
16 The remaining studies that were submitted
17 as "supportive" studies, it did not mean that, in
18 our opinion, any of these studies could not also
19 have been pivotal from the point of view of their
20 design being adequate and well-controlled, having
21 case-report forms, electronic databases, and so on,
22 but it was more a question--in some instances, the
23 study centers were not available anymore or the
24 practitioners who were there weren't available.
25 So the three studies that we
207
1 identified--and we identified those from the very
2 beginning, the PRAMA and Paille study, and we added
3 the Belgian-French study, the Pelc II study. These
4 were always going to be--or at least the first two
5 were always going to be part of our pivotal
6 database.
7 We did not submit, in the NDA, the U.S.
8 study as a pivotal study and we really think it is
9 misconstruing to say that we thought this was a
10 pivotal study. We didn't. We feel as interested
11 as the committee and the FDA in understanding why
12 the results weren't the same as the European
13 studies for the ITT population, but we think we
14 have done a good job in terms of trying to get an
15 interpretation on a subgroup that could really
16 benefit from the drug.
17 DR. OREN: Given that the European studies
18 are a key to our discussion of efficacy, could you
19 also clarify further or tell us more about the data
20 structure in those studies and the capacity of
21 those specific studies to provide valid endpoints
22 for us.
23 DR. GOODMAN: If I could, I would like to
24 ask Dr. Cook to address that point with Dr. Mason's
25 help, perhaps.
208
1 DR. G. COOK: The European studies had
2 assessments at specific time intervals. My
3 understanding is that those assessments would be
4 considered sufficient to identify departures from
5 abstinence, that, if a patient had a departure from
6 abstinence, it would be likely to be a major
7 departure and, through the various reporting
8 mechanisms, one would have been able to have
9 captured such a departure.
10 Now, that simply means that when you focus
11 on an abstinence-oriented endpoint, things are
12 fairly straightforward, whether it is complete
13 abstinence throughout the time period in the study
14 or time to first departure from abstinence or even
15 the number of assessments in which abstinence was
16 reported.
17 Certainly, the FDA has correctly
18 identified some difficulty in a calculation of
19 number of days with abstinence because that
20 involves some assumption about the time interval
21 between the assessments. I think the spirit of the
22 sponsor's categorization of all days subsequent to
23 an assessment of nonabstinence as drinking days was
24 simply based on the principle that if a patient had
25 a departure from abstinence, they would be
209
1 considered a drinker until the data structure
2 proved that they were no longer a drinker.
3 Again, that was probably based on the
4 philosophy that a departure from abstinence is not
5 something that just occurs for a few hours or a day
6 or two but that it actually is a total return to
7 the alcoholism for which they originally were being
8 cared for.
9 Whether that assumption is right or wrong,
10 I can't really comment on. I was just trying to
11 give some clarification as to why the sponsor,
12 potentially when they developed this strategy--why
13 they basically called all days after a nonabstinent
14 day a drinking day following essentially a
15 last-observation-carried-forward principle.
16 Perhaps my colleagues here can comment on it
17 further. But, regardless of how you choose to deal
18 with that intervening interval, I believe that
19 abstinence was accurately characterized by the data
20 structure because, again, I think my colleagues can
21 reinforce the point that if a patient had a
22 nonabstinent episode, that data structure was
23 probably adequate to capture it.
24 So I would like Dr. Mann and Dr. Mason,
25 perhaps, to comment on these points further.
210
1 DR. MASON: In terms of the importance of
2 a slip or a drinking episode, as I had mentioned
3 earlier, one of the diagnostic criteria for alcohol
4 dependence is going on to--when the person with
5 this disorder, one of the ways they are
6 characterized is by their going on to drink much
7 more than they originally intended.
8 They may go to the wedding reception
9 planning on having just one drink and wake up a
10 case of beer later. That is one of the hallmarks
11 of the disease. All of the intervals that were
12 used as assessment intervals in the European trials
13 were of sufficient duration as demonstrated by
14 Sobell and others working as methodologists in the
15 area of alcohol dependence. They were of
16 sufficient duration to capture these important
17 episodes of abstinence and nonabstinence and long
18 enough to capture episodes of infrequent drinkers.
19 If you have a very quick rating period, it
20 is possible that you would miss a drinker because
21 the drinking just hadn't occurred in a narrow
22 interval. You do need an interval of sufficient
23 time to capture the infrequent drinkers who have
24 more of the binge-type pattern.
25 A final point that I would like to make
211
1 about the duration of the intervals that the
2 European data collection used and the method in
3 which the drinking data were collected in Europe is
4 how closely it follows U.S. clinical practice. I
5 believe that, given how the methods and the
6 intervals follow clinical practice, and the
7 benefits shown with acamprosate in this type of
8 setting and under this level of inquiry will
9 likewise benefit U.S. patients with alcohol
10 dependence that was diagnosed under exactly the
11 same set of criteria as those patients with alcohol
12 dependence in Europe.
13 DR. MANN: I certainly agree with what has
14 been said about a slip and how a slip is a short
15 return to drinking in general mounts up to what is
16 a full-blown relapse in 80 to 90 to 95 percent.
17 So, in taking into account a slip and counting it
18 as a relapse until the next visit, I think that was
19 the most conservative and the most valid way of
20 looking at these data.
21 I would also like to mention one more
22 point. The German study, the PRAMA study, was
23 published in the Archives of General Psychiatry in
24 1996 and that would say something about the
25 validity of the self reports using gamma GT,
212
1 figures that we have not heard yet today.
2 There we state that between 81 and 100
3 percent of the patients who self-reported relapses
4 had higher gamma GT levels in both groups, so there
5 was no difference between both groups, and also
6 that the gamma GT values above the normal reference
7 range also corresponded with the number of patients
8 who had had relapses, again in both groups.
9 So I think there is some solid evidence
10 that these self reports are validated by external
11 sources.
12 DR. OREN: We will now turn the discussion
13 over to the committee for us to discuss amongst
14 ourselves. Dr. Titus reminds me that we can also
15 feel free to ask the FDA, ask the sponsor,
16 questions that are of relevance to our discussion
17 to further us along.
18 Obviously, the three questions we have
19 been charged with are all interrelated with each
20 other but, perhaps, we can start with one and try
21 and focus on one and move towards the other. The
22 first one is how can we reconcile discrepant
23 results between the older European studies and the
24 more recently concluded American study?
25 Dr. Hamer?
213
1 DR. HAMER: First of all, I want to say
2 that I am less than impressed by the argument that
3 the assessment methodology in the European trials
4 followed closely the clinical practice in the
5 United States. That seems to me to be an analogous
6 argument for not using the Hamilton Depression
7 Scale in our depression studies because, after all,
8 in clinical practice, we don't use the Hamilton
9 Depression Scale to assess our patients.
10 DR. OREN: Dr. Fuller?
11 DR. FULLER: My comment is somewhat
12 related and it was already made earlier by two
13 individuals. One was Dr. Hamer. That is the issue
14 that if you do several clinical trials, you will
15 get discrepant results. One, perhaps, is just by
16 chance. Another could be different methodologies.
17 But we have already mentioned the
18 depression studies where this is not an uncommon
19 occurrence, at least as reported in Science last
20 year. Even with effective therapies, you will get
21 some studies where the medication is no better than
22 the placebo.
23 But the comment I wanted to make was more
24 of a historical nature and that has to do with
25 aspirin for preventing myocardial infarction in
214
1 people who have myocardial infarction. This is
2 considered very important by cardiologists and
3 groups such as Medicare who pays for healthcare.
4 Yet, there was a similar situation where there were
5 two positive studies and then there was a large
6 negative study that involved 2000 individuals.
7 Then there was a fourth study.
8 On the basis of three positive studies,
9 one negative study, people undertook metaanalysis
10 and it has become faith that people who have had a
11 myocardial infarction ought to have aspirin and
12 even those who don't should have it. I just wanted
13 to bring that historical vignette in.
14 What I was leading up to was this
15 sometimes happens, these discrepant results.
16 Others here may have insight into why they happen,
17 but we may not be able to reconcile the discrepant
18 results. But they do occur.
19 DR. OREN: Dr. Winokur?
20 DR. WINOKUR: To begin to address the
21 issue of the discrepant results, we certainly heard
22 a lot of discussion this morning about some
23 important differences in the populations included.
24 One important point that was mentioned and
25 acknowledged by the FDA is the request to broaden
215
1 the scope of patients including polysubstance use
2 for safety assessment, and that clearly may have
3 changed the composition of the population
4 considerably.
5 But picking up on a comment that Dr.
6 O'Brien made, I also wanted to raise the question
7 as to whether there may be a change in the
8 treatment of alcohol dependence and whether this
9 changed the nature of patients available for
10 studies that occurred earlier in the European
11 studies which were, as we have mentioned, over a
12 decade ago with the more recent studies.
13 Again, the other major difference that we
14 are really grappling with is the issue which was
15 unexpected of the substantial number of patients in
16 the U.S. study who were not abstinent at the time
17 of start of treatment. I know we heard from the
18 FDA that they are willing to accept studies from
19 Europe as a basis for approval but I wonder if
20 there is a reason to discuss whether situations may
21 have changed not necessarily with the illness but
22 the environment in which people are not carrying
23 this illness and are being treated such that the
24 population being studied more recently really
25 represents a different cross-section.
216
1 DR. OREN: Do you want to say more? I
2 think that is an intriguing thought.
3 DR. WINOKUR: I was really hoping to get
4 some input from people that really work in the
5 field with this population which I certainly don't.
6 DR. OREN: Dr. O'Brien?
7 DR. O'BRIEN: Just to continue on the
8 theme of Dr. Winokur, actually is it what Dr.
9 Winchell said. There are clear differences that,
10 in the populations, in terms of--first of all, the
11 environment, the availability of detoxification is
12 a major difference. The number of people who
13 started off not being detoxified. That is a big
14 difference in all the addicting drugs that we
15 study.
16 The coincidence of other kinds of
17 substance abuse at the same time makes for a more
18 heterogenous population. We haven't said much
19 about comorbid other diagnoses but we know that
20 there is a very high comorbidity of anxiety
21 disorders and affective disorders in alcoholics.
22 That has tended to vary both in different countries
23 and in different sites.
24 For example, some questions were raised
25 earlier about what is the percentage of alcoholics
217
1 who have substance abuse or who have one thing or
2 another. It really depends on whether you are
3 talking about a community program, a V.A. program,
4 an HMO, a private program. Every environment that
5 you go to is different.
6 So you have all of these environmental
7 factors. Of course, the time. For example, if you
8 did these studies in Germany or France today, you
9 might find a lot more comorbid substance abuse
10 because I believe that there are a lot more street
11 drugs available over there now.
12 But, in addition to all of these factors,
13 you have the biological differences in alcoholism.
14 We all know that there are different ways of
15 categorizing alcohol and the current ones are Type
16 1, Type 2, A and B. But none of these really
17 capture what are probably endophenotypes that,
18 among people who may use the same amount of grams
19 of alcohol per week but they are biologically very
20 different.
21 For example, if we give them alcohol in
22 the laboratory, one difference that is
23 extraordinary is the fact that some people get a
24 huge increase in plasma beta endorphin and other
25 people don't. They also get a different response.
218
1 It is either activation from alcohol or sedation
2 from alcohol.
3 When we give them the other drug that has
4 been mentioned here, naltrexone, some people, it is
5 just life-saving in the sense that they say that,
6 gee, it has really turned my life around and they
7 get a tremendous benefit from it and, if we stop
8 it, they relapse to alcoholism. So there is no
9 doubt in the mind of the patient and the person
10 treating the patient that the drug is active.
11 But, on the other hand, there are other
12 patients for whom you give the drug and there is no
13 benefit whatsoever even though, according to the
14 usual classification of alcoholism, they might be
15 identical. So we haven't come to the point in
16 alcoholism where we can make a diagnosis like, for
17 example, with anemia. We can take two people with
18 an hematocrit of 30 percent but we know that, by
19 doing hemoglobin electrophoresis, they may have
20 totally different kinds of anemia and you would
21 treat them totally differently even though their
22 symptoms are very similar.
23 We maybe someday--I hope, someday--will be
24 able to do that with alcoholism but to have
25 complete lack of divergence across clinical trials
219
1 would be totally unreasonable today since we are
2 lumping together people who are very heterogeneous
3 not only according to the environment things that
4 Dr. Winokur brought up but also according to the
5 biology of the illness.
6 DR. OREN: Dr. Rudorfer?
7 DR. RUDORFER: Just to follow up Dr.
8 O'Brien's comments, in addition to some of these
9 cross-sectional issues, I just want to remind us
10 about the longitudinal aspect of this disorder.
11 Several of us have made references to mood
12 disorders, a similar kind of chronic relapsing
13 recurrent disease.
14 It seems to me that, just to kind of
15 restate something we have been saying from a
16 different perspective, there are certainly
17 different phases of the illness of alcoholism and I
18 fear that sometimes those have gotten lumped
19 together here today just in terms of talking about
20 treatment of alcoholism.
21 The issue with the percent of patients
22 abstinent at baseline I think is important in terms
23 of considering the phase of the illness so that the
24 European data really point to efficacy in the
25 prevention of relapse in patients who are already
220
1 abstinent, and not just already abstinent but
2 abstinent following an inpatient detoxification.
3 That is a particular stage of this illness and many
4 people may go through that multiple times during
5 their lifetime or not at all but to intervene at
6 that particular point, I think, is simply not the
7 same as intervening at another point.
8 So, to a certain extent, I see a certain
9 amount of apples and oranges in the European and
10 the U.S. trials.
11 DR. OREN: Dr. Hamer?
12 DR. HAMER: I think it is unfortunate that
13 the U.S. trial was almost an effectiveness study
14 rather than an efficacy study because what was
15 probably needed was an additional efficacy study in
16 the U.S. In terms of the decision we are being
17 asked to make, I think that, regardless of the way
18 that the sponsor presented the data and regardless
19 of the way we listen to it, it is clear from the
20 FDA's charge and from the things that have been
21 said elsewhere that, except for the issue of trying
22 to reconcile what happened in the U.S. study versus
23 the European studies, that the decision to approve
24 and probably, thus, most of our deliberations to
25 that part of addressing efficacy ought to be based
221
1 on the European studies, and the U.S. study ought
2 to be viewed as simply an additional failed study
3 and we should attach no more and no less weight to
4 that then we would in similar situations.
5 Having said that, the data the sponsor
6 presented showing efficacy of a sort in the U.S.
7 study depended upon what might appear to be a
8 carefully crafted set of covariates figured into
9 the analysis post hoc. Evidence that those
10 covariates are useful and meaningful and, in fact,
11 mean something in the course of the U.S. study
12 would be useful to us.
13 One way to address that might be to take
14 those same covariates or ones as similar as you can
15 obtain in your database and apply them in the
16 European data and see if they improve the effect
17 size. I wonder if you have done anything like
18 that.
19 DR. LEHERT: My name is Philip Lehert from
20 the University of Brussels and the World Health
21 Organization. I have examined, as a third party,
22 the whole database coming from acamprosate from the
23 European and the American data. I have done
24 exactly what you said.
25 I have examined 4,500 patients on the
222
1 basis of initial motivation, whether they drink or
2 not, and ten or fifteen different covariates. I
3 found exactly the same covariates in European as in
4 the United States. In using the same covariates on
5 the 4,500 patients in my model, I just used the
6 interaction between the United States, yes or no,
7 and the treatment.
8 I found a significant effect of these five
9 covariates and no significant effect of the
10 interaction. This is just telling you that what I
11 have done is justification of these five covariates
12 all around the world.
13 DR. HAMER: Although, depending on which
14 model we are talking about, there were either six
15 or seven covariates used in the U.S. trials.
16 DR. OREN: Could you identify those
17 covariates?
18 DR. LEHERT: Yes. The first I have,
19 unfortunately, not slides of that but I would just
20 like to say that this would belong to part of the
21 dossier. The first was whether or not the patient
22 was motivated. I would like to stress that
23 motivation was part of the European data but just I
24 had to take this data on the CRFs, themselves.
25 The second was the most important variable
223
1 I found for all the five. This was whether or not
2 the patient was drinking at baseline. I would like
3 to stress that the FDA has done the same analysis
4 of the American study but just on the seven first
5 days.
6 I did it on the basis of time-line
7 follow-back for Day 0, just Day 0. In other words,
8 I am able to look at all the patients that were
9 drinking at baseline and I found this very
10 surprising and very interesting medically speaking
11 results that the interaction of acamprosate and
12 abstinence at baseline was more important than the
13 acamprosate main effect only. This means that
14 before being treated by acamprosate, a patient must
15 be good willing to heal and not drink at baseline.
16 My first impression in the United States
17 data is that when I just look at those patients who
18 are not drinking at baseline, I found different
19 results in line with the European results.
20 I have a very last thing to say which is
21 the four other main effects were medication
22 compliance, and I would like to stress that it is
23 not compliance during the trial but the compliance
24 measure at the beginning of the trial and it is
25 just at the first three days we had this question.
224
1 My question was to know whether or not, in
2 using the compliance in the beginning, would should
3 have some image of the motivation because you know,
4 in the European data, I had no motivation of the
5 patient. In other words, I had to find another way
6 of measuring the motivation of the patient and I
7 found that in two things.
8 The first was that whether or not they
9 were drinking at baseline and second if they were
10 good willing to be compliant for the three first
11 days. That is what I found. And I finish in
12 saying that a moderate baseline I will call
13 dependency severity I suppose that everyone can
14 understand that the severity of the illness can be
15 of some importance in the predictive model. At the
16 end, just living with a partner and a child was the
17 thought.
18 I am happy to tell you that on my 4,500
19 patients, I was able to collect more than 35
20 percent of the whole variance which makes that my
21 model is somewhat explanatory, something that never
22 happens even in the World Health Organization in my
23 predictive models. I was very happy to have that.
24 And, at the end, what I was able to see is that
25 there was no interaction when I put that out
225
1 between the U.S. and the non-U.S. data.
2 In other words, there was no interaction
3 between the country, the trials and the product,
4 itself. In other words, my selection of my four
5 different endpoints was probably favorable for
6 explaining exactly. This is what we call a
7 metaanalysis based on individual patient data.
8 Thank you.
9 DR. OREN: Dr. McCormick.
10 DR. McCORMICK: I would just like to
11 caution the committee that these are not analyses
12 that we have had the opportunity to review and to
13 comment on. In fact, we haven't seen most of the
14 sixteen trials in detail that you have used in this
15 reanalysis. So I would caution the committee not
16 to rely too heavily on something that we have not
17 had the opportunity to review carefully.
18 DR. OREN: As committee members, we are
19 also at the same level of ignorance as far as
20 awareness.
21 Dr. Hughes?
22 DR. HUGHES: Just a quick yes/no question.
23 When you did your analysis, did you look at just
24 abstinence as a covariate and get an
25 interaction--not the full four, just abstinence,
226
1 because clinically no one is going to say, well, if
2 you got a, b, c and d, I will give you the drug.
3 The most we can get is, perhaps, one thing. So,
4 with just abstinence did you find an interaction?
5 DR. LEHERT: I just used the fact on the
6 TLFP that I had abstinence in drinks every day and
7 I just looked at Day 0 and the very beginning of
8 Day 1. Then I repeat. I apologize to come back to
9 the study, if you allow me that, that this variable
10 was by far the most important predictor of success.
11 I think it was so important that I put that.
12 DR. HUGHES: But I am just asking if you
13 just had the model with just abstinence as the only
14 other thing in the model, did you show an
15 interaction of abstinence with treatment
16 assignment?
17 DR. LEHERT: Yes; I did
18 DR. HUGHES: Thank you.
19 DR. WINCHELL: Just to clarify, this was
20 the European database combined with the American
21 database that you analyzed this way?
22 DR. LEHERT: I analyzed in a metaanalysis
23 file all the data together including the American
24 study. That's right.
25 DR. WINCHELL: So the only subjects who
227
1 were not abstinent at baseline in your 4,500
2 patients were from the American study and then a
3 few in the U.K. study; correct?
4 DR. LEHERT: Yes; that's correct.
5 LIPHA: Just as a point of clarification,
6 that was submitted as a part of the integrated
7 summary of efficacy.
8 DR. WANG: Can I just add? This is Sue
9 Jane Wang from the FDA. In the analysis for the
10 U.S. study when just the abstinence goal was
11 included in addition to treatment in the center
12 that was included the model, I get the p-value of
13 0.431 of the medium dose compared to placebo. But
14 this is just for the U.S. study.
15 In other words, if you adjust for that
16 prognostic covariate, I do not see a treatment for
17 the medium dose.
18 DR. OREN: Dr. Rudorfer?
19 DR. RUDORFER: A question for the sponsor.
20 We have been discussing today studies that lasted
21 from six to twelve months. I am wondering if you
22 had any secondary measures in terms of function of
23 quality of life that would help us understand the
24 efficacy data better?
25 DR. GOODMAN: We did not include anything
228
1 regarding quality of life in the NDA. I know that
2 Dr. Lehert has done such an analysis of the
3 European data but it has not been submitted with
4 the NDA.
5 DR. OREN: Dr. Hamer?
6 DR. HAMER: I just wanted to confirm; in
7 the metaanalysis of individual patients that you
8 did, you used all the U.S. and European subjects.
9 So what you don't have is confirmation in the
10 European data alone that the same predictor--that
11 is, abstinence--is predictive in the European data
12 alone as it was or was not in the American data and
13 that also, since basically you had all abstinent
14 patients in the European data, that variable really
15 is largely confounded with the European versus
16 American studies; right--since half the U.S.
17 patients were not abstinent and none of the
18 European patients were--excuse me; half of the
19 American patients were not abstinent and none of
20 the European patients were not abstinent.
21 DR. LEHERT: The U.K. patients are
22 included into this data file metaanalysis and I
23 think, and I presume, that everybody's view of
24 statistics will assume that it would be doubtful to
25 make at least an analysis on only U.K. What I did
229
1 was that every time I assessed my model, I used a
2 defined protocol for analyzing the interaction of
3 the first order and then every time this
4 interaction was found, I included it in the model.
5 What I found was that only the interaction
6 between abstinence and the treatment was present in
7 my data. But I have done that exactly as you said.
8 DR. OREN: Dr. Hughes?
9 DR. HUGHES: Dr. Hamer, I am thinking very
10 differently than you here. The FDA said that when
11 they had abstinence, they didn't find anything. So
12 if he is finding in the full dataset, it must be a
13 whopping effect in the U.K. to swamp out the lack
14 of interaction in the U.S. Am I thinking right
15 here?
16 DR. HAMER: Or suppose there was no
17 abstinence effect in the U.S. study, that half the
18 patients in the U.S. were abstinent and also in the
19 U.S. study, abstinence didn't make a difference and
20 also in the U.S. study, we didn't show much of an
21 acamprosate effect. In the European studies, let's
22 suppose exclusive of the British study because that
23 is a small portion of the patients they have there,
24 everyone was abstinent and there was an effect,
25 therefore the difference you find in sort an
230
1 acamprosate effect versus a nonacamprosate effect
2 is fairly confounded with the U.S. versus European
3 studies and also fairly confounded with abstinent
4 nor nonabstinent. So that is not surprising.
5 I don't think I sort of asked my question
6 adequately. What I would have liked to have seen,
7 since they presented three European studies as part
8 of the NDA, would have been an independent
9 confirmation in the data from those three studies
10 alone, not including the U.S. data and not
11 including any of the other European data, that the
12 same set of covariates showed prediction in those
13 data as well, in the same way as they did in the
14 U.S. data.
15 DR. G. COOK: I think I understand what
16 your question is. I think that those analyses have
17 not been done. I think they mainly have not been
18 done because the direct analyses of the European
19 studies, regardless of covariate adjustment, do,
20 indeed, show significant results. The analyses
21 that the sponsor has done with the U.S. study are
22 largely explanatory. They are not being done to
23 prove anything because they couldn't prove anything
24 even if they found something that looked
25 attractive.
231
1 They are simply an attempt to see whether
2 or not they can identify trends that seem to be
3 consistent with the findings in the U.S. study. A
4 rather key part to the analyses they did along
5 those lines is to hone in on the motivated group
6 and the motivated group that you have to work with
7 for that purpose is the group that is motivated to
8 be abstinent in the strictest sense.
9 You also have to do the analysis that, in
10 the denominator, uses all days, if people dropped
11 out for an alcoholism-related reason and use days
12 up to time of discontinuation if they dropped out
13 for some other reason and that other reason was
14 considered credible.
15 But these analyses are more to identify
16 trends. They are not necessarily analyses that are
17 intended to produce attractive p-values. You don't
18 get attractive p-values that are durable in the
19 U.S. study. You can find suggestions in the U.S.
20 study that some of you may find reassuring but you
21 need to make your decision on the basis of your
22 confidence in the efficacy shown in the European
23 studies with whatever reassurance you are finding
24 from the U.S. study, recognizing that finding that
25 reassurance may be hard.
232
1 DR. OREN: Dr. Schatzberg?
2 DR. SCHATZBERG: This bears on that. This
3 is for the sponsor. If you look at the dropout
4 rates on the U.S. study, the dropout rates on
5 active drug are pretty high, particularly on the
6 2000 milligram per day dose. They run about 60
7 percent. I am just wondering how you reconcile
8 that kind of dropout with Dr. Mann's comment about
9 the PRAMA study, the German study, in which staying
10 in was seen as a good thing.
11 What kind of assurance can you have that
12 this doesn't mean that this isn't really a kind of
13 a really very fallible, very flawed study where
14 nobody stays in and 60 percent of the patients
15 dropping out. You can't have it both ways in the
16 argument. If you are the sponsor, you can't say,
17 yeah, people stayed in, it's great and then, in a
18 very large-scale trial, you have a very, very poor
19 completion rate.
20 So I don't know how you reconcile the two
21 arguments in the same presentation.
22 DR. GOODMAN: I don't plan to answer that
23 directly but I think that we were trying to
24 demonstrate, and again, just from an interpretation
25 as to how to explain our results in the ITT
233
1 population, I believe what Dr. Mason was trying to
2 do when she reviewed the demographics was to show
3 you that, collectively, we considered this 2-gram
4 group to be somewhat disadvantaged in a variety of
5 demographic measures, or baseline measures, relate
6 to drinking.
7 Barbara, I don't know if you want to say
8 anything more.
9 DR. MASON: It wasn't just their
10 disadvantage in relation to drinking. It was the
11 fact that they also had fewer psychosocial supports
12 like full-time employment, living with someone.
13 These are all aspects of rootedness and structure
14 that contribute to stability and staying in
15 treatment. Also, in general, in terms of the high
16 rate of dropouts, that is something that has been
17 demonstrated very nicely by the group at the
18 University of Connecticut where they looked at
19 dropout rates across clinical trials involving the
20 addictions, primarily illicit drug use, relative to
21 dropout rates involving clinical trials for other
22 psychiatric disorders.
23 The difference in the rate of dropouts
24 were very marked, particularly as one gets into
25 illicit substance use. So I believe that that
234
1 probably also colored the dropout rates of the U.S.
2 study that was so characterized by illicit drug
3 use.
4 DR. OREN: Dr. Leon?
5 DR. LEON: Let me follow up on what Dr.
6 Mason just said. The slide that she showed, each
7 of those differences at baseline looks very
8 trivial, 2 or 3 percent. Certainly, none of them
9 were statistically significant so I don't think we
10 should overstate the importance of that. They are
11 on the slides on Page 8 of your handout for anyone
12 that wants to see.
13 I want to say a couple of other things.
14 The intent-to-treat principle was referred to in
15 the analysis. The sponsor referred to that for the
16 pivotal trials. It is my understanding, though,
17 the that intent to treat was applied in an
18 unconventional way where the last observation was
19 carried forward, imputed for all data after
20 subjects dropped out of the trial.
21 In other words, the treatment and
22 assessment were very tightly linked. As soon as
23 someone stopped receiving treatment, they stopped
24 being assessed. Is that correct? Before I get the
25 answer, I look at the intent-to-treat principle to
235
1 be more tightly interpreted, to mean that, whether
2 or not somebody is receiving treatment, the
3 assessments are continued for the duration of the
4 trial.
5 DR. G. COOK: So that would mean you would
6 only be confident in a trial that had zero
7 dropouts.
8 DR. LEON: No. I just wouldn't call it an
9 intent-to-treat analysis. I wouldn't call what
10 they refer to as an intent to treat invoking the
11 intent-to-treat principle. They are imputing data
12 with the last observation carried forward.
13 DR. G. COOK: So you are saying that you
14 can only do intent to treat when there are zero
15 dropouts.
16 DR. LEON: No; that is not what I am
17 saying. That is what you are saying.
18 DR. G. COOK: But if what they did as an
19 analysis of all randomized patients is not an
20 intent-to-treat analysis, then it can only fail to
21 be not intent to treat because it imputed a failure
22 status to a dropout.
23 What it basically did was it had a certain
24 number of patients complete and, in the European
25 trials, you would have had a status of the patient
236
1 at the time of completion. The patients who
2 dropped out were basically managed as treatment
3 failures.
4 Now, the FDA did analysis in which they
5 managed those dropouts in other ways. There was
6 also an attempt to look at time to first departure
7 from abstinence as well. That was the
8 time-to-event analysis. That tried to deal with
9 the data. But, to avoid a semantic difficulty,
10 whatever the sponsor called intent to treat, I
11 believe was simply referring to all randomized
12 patients or all randomized patients with a few
13 exceptions who may not have taken at least on dose
14 of treatment. But that was only a small number, I
15 think, in Dr. Mann's presentation.
16 DR. LEON: Just so I understand this, this
17 was all randomized subjects were included and
18 assessed until they dropped out but none, or very
19 few, were assessed after they stopped taking their
20 treatment; is that correct?
21 DR. G. COOK: That's correct. There is
22 not a retrieved dropout.
23 DR. LEON: Although an alternative
24 strategy, assessment strategy, would be to continue
25 to assess the patients after they stop taking their
237
1 drug.
2 DR. G. COOK: Yes. And that is very much
3 recommended in today's environment although, again,
4 my understanding from Dr. Mann and others is that a
5 patient who drops out when they are being treated
6 for alcoholism is a patient who is very, very
7 likely to relapse, that these patients are very
8 fragile and, to some extent, dropping out is almost
9 tantamount to treatment failure.
10 Perhaps Dr. Mann would want to comment on
11 that further, or Dr. Mason.
12 DR. MASON: Andy, a point I would just
13 like to make in dealing with this population is
14 that once they are gone, they are really gone. It
15 is very hard to track them after they have lost
16 control of drinking. That is why this type of
17 intervention is so critically important just to
18 keep them involved in treatment.
19 Then, if there is a relapse, as long as
20 they are involved, as long as they remain engaged
21 for whatever reason, you can get them through the
22 relapse. I believe that the label for acamprosate
23 says to continue administering during a relapse.
24 But in a clinical trial involving
25 outpatients with alcohol dependence, once they are
238
1 gone--it is not like where you can very practically
2 say, you are going to continue research assessments
3 even though they have left the treatment arm of
4 involvement. It just tends to go when you have
5 someone really lose control in that way.
6 DR. OREN: Dr. Cook
7 DR. COOK: This is for the FDA. Did you
8 find evidence that they had documented how they
9 were going to handle failures in the
10 analysis--predefined, of course?
11 DR. WANG: For the European trials
12 DR. COOK: Maybe I can make a comment in
13 terms of how I am thinking about the questions.
14 The U.S. trial was failed.
15 DR. WANG: Also, the algorithm was
16 prespecified
17 DR. COOK: Pardon me? So now my question
18 is about the European trials because what I am
19 really trying to focus on is do we have evidence
20 for more than one adequately conducted controlled
21 trial for efficacy? The U.S. trial is not going to
22 be it. The sponsor acknowledges that. But I hear
23 questions about the three European trials.
24 I keep coming back to the point of
25 predefined analysis endpoints, et cetera, and how
239
1 failures are handled.
2 DR. WANG: My understanding is, for the
3 European trials, the definitions of dropouts, who
4 they are going to evaluate, as I showed in all the
5 slides, I distinguished between dropout as is and
6 as relapsed.
7 DR. COOK: So my question is what did the
8 sponsor predefine as the way they were going to
9 handle dropouts?
10 DR. WANG: I guess maybe we can go trial
11 by trial. The Pelc II trial was a three-month
12 study. Because we weren't very sure about those
13 imputations for the CAD data, cumulative abstinence
14 duration, the way to analyze these data, we can
15 only say the way they do the imputation on the
16 dropout patients, in some trials, they used the
17 worst-case analysis, worst-case here, I mean they
18 would impute all the dropout patients as patients
19 who relapsed.
20 But they don't do this consistently across
21 the three trials.
22 DR. COOK: Let me clarify because I think
23 we are getting into a little bit of an metaanalysis
24 of all the studies instead of coming back to the
25 principle that Dr. Leon pointed out that, to me, is
240
1 what we have to adhere to. If the three are
2 slightly different but within reason, I want to
3 know what was the analysis they prespecified.
4 Did they write that down? Is that a
5 document that we can verify and did their primary
6 specified analysis show a difference? We have
7 gotten confused. One page would be more helpful
8 than hundreds.
9 DR. WANG: For the three European trials,
10 we really don't know. That is why we are
11 struggling with presenting two ways of dropout as
12 is versus as relapsed. We have trouble with the
13 definition of what is the primary efficacy outcome.
14 It was not really stated.
15 DR. OREN: I would like to, at this point,
16 use this as a segue in our discussion to move away
17 from the first question, which was how can the
18 discrepant results be reconciled and to summarize
19 that.
20 We have heard at least that there may have
21 been different outcome endpoints between the
22 American study and the European studies. There are
23 certainly different levels of rigor. Randomness
24 may play a role and just this happened to be an
25 unlucky American study, different times, ten years
241
1 ago versus two years ago, different populations,
2 European versus American, different populations as
3 far as comorbid substance abuse, whether people
4 were drinking at the time of entering the study.
5 We have just heard about a metaanalysis
6 that suggests that maybe they can be easily
7 reconciled. I sort of feel like it is the old
8 Perry Mason show where a surprise witness comes in
9 at the end except in this case I am no judge. But
10 we don't have the full evidence to be able to
11 consider it at this point.
12 But this is, I think, at least the
13 background. At this point, this might be a good
14 time to move to the central question of, given the
15 results that we have seen today, and it seems
16 predominantly the European studies that we are
17 interested in, is there sufficient evidence of the
18 efficacy of acamprosate in the treatment of
19 alcoholism to warrant approval.
20 Again, we will take a vote whether to
21 recommend on the efficacy question, to make a
22 recommendation to FDA on how to act in that regard.
23 In that vote, I will go person-by-person through
24 the entire committee asking everybody to register
25 their vote, yes, no or abstain.
242
1 But, before that, we have open time for
2 discussion and I would certainly invite everyone,
3 in the course of this discussion, to make your
4 viewpoint known if you like.
5 Dr. Fuller?
6 DR. FULLER: I think my question bridges
7 both Question 1 and Question 2 in that we were just
8 discussing whether there were predetermined
9 endpoints in the European studies. I can be
10 corrected if I am wrong, but when I read this
11 document, I thought two of the three European
12 studies did have predetermined endpoints. I
13 believe--I think, analysis, but the predetermined
14 endpoints, as I read them was in the Pelc study was
15 sustained abstinence and in I will call it the
16 German study was time to first drink. I think that
17 is what they had decided initially to use as
18 endpoints.
19 Then I believe that there was also an
20 endpoint for all three studies added on slightly
21 later, the cumulative abstinence days. I think I
22 am speaking correctly.
23 DR. LEON: I am working from this
24 document. I will show you the page numbers.
25 DR. WINCHELL: Which document?
243
1 DR. LEON: The FDA background document.
2 If you turn to page 32 of the medical record from
3 Dr. Winchell's report, the evaluation of endpoints,
4 Section 5314, the prespecified main criterion of
5 judgment listed in the protocol was, "the
6 consumption of alcohol, no a prior strategy for
7 transforming the data collected into an overall
8 assessment of alcohol consumption was identified."
9 Also, on that page, as long as we are on
10 that page, there is no explicit data-analysis plan.
11 That is the next big paragraph down.
12 If we turn to the Paille study, Page 13 of
13 the statistics in the FDA document, the last
14 paragraph on Page 13, the first sentence, says that
15 no statistical-analysis plan was included here and
16 the protocol-dependent variable is also on that
17 page, the primary efficacy endpoint is here. The
18 number of abstinent days is right above that
19 paragraph, but this is not the one that was used in
20 the analyses that were presented.
21 As long as we are on this trial, I do want
22 to quote from the sponsor's report that there was
23 not a significant difference between 1332
24 milligrams and placebo. I think that has been lost
25 in the discussion today. In the Paille study, the
244
1 sponsor's report said there was not a significant
2 difference between placebo and the 1300 milligrams.
3 If you want to see where I got that, that
4 is in FDA report, Page 18, of the statistics
5 report.
6 DR. OREN: Although, since the protocol is
7 for approval for 2000, is that still a problem?
8 DR. LEON: Oh; if we are going to ignore
9 all studies that didn't test 2000, we would knock
10 out some other European data, wouldn't we? We
11 would knock out a third of the data from Pelc and
12 what else?
13 The other dependent variable, though, as
14 long as we are going through these, in PRAMA, was
15 time to relapse. That was defined on Page 61 of
16 the medical record from the FDA. That was time to
17 relapse and that was the day on which alcohol
18 consumption started again.
19 So that is my point of clarification on
20 the dependent variable.
21 DR. G. COOK: I think you are identifying
22 some of the same kinds of considerations that the
23 FDA reviewers identified in the course of their
24 review which is that studies that were launched in
25 the late 1980s and the early 1990s did not have
245
1 detailed statistical statements in their protocols
2 and they may not have had detailed statistical
3 analysis plans that were formally written prior to
4 unblinding.
5 Because of that, it becomes important for
6 analyses of the data structures that those studies
7 produced to be relatively consistent and robust.
8 So that is why it was somewhat important for the
9 FDA, in their reanalyses under any number of
10 conventions, to find similar significant results to
11 what the sponsor found in their analyses. It is
12 much more critical that the majority of analyses
13 agree with one another in terms of p-values below
14 0.05 when you do not have detailed plans that are
15 identified up front.
16 That is why the robustness from both the
17 FDA analyses as well as the sponsor looking at
18 several things all pointing in the same direction
19 was something that had some discussion.
20 DR. OREN: Dr. McCormick?
21 DR. McCORMICK: I tend to agree with Dr.
22 Cook in his assessment of the quality of
23 prospective strategies in some of the older
24 studies. I think, in our frustration when we
25 reviewed these studies, of not having carefully
246
1 laid out primary endpoints and statistical analyses
2 plans and so forth, led us to take probably the
3 most rigorous approach we possibly could take.
4 So we basically looked at these trials
5 with the perspective of what is the highest bar we
6 could set for these studies and it was complete
7 abstinence. We felt that the studies made it on
8 that criteria.
9 Our discomfort, as I mentioned this
10 morning, is--I think we have almost moved past this
11 problem of not having the prospective strategies
12 before us and that is really dealing with the issue
13 of this imputed data. Do we believe it or not? Is
14 it really credible? Three months of really no
15 ascertainment, can we know what really happened or
16 not?
17 If I were to summarize the crux of our
18 discomfort, it has to be that.
19 DR. MANN: That is something I understand.
20 I think, in looking back at these in our early
21 days, we have the same kind of discomfort. But,
22 fortunately, we have also other data, the ones that
23 were shown by your statistician, which is
24 abstinence rate per visit. Only one day, and you
25 take all the information that you can get and you
247
1 say someone is abstinent or is not abstinent.
2 You are not computing back or forth or
3 anything. You just say, today is abstinent or nor
4 abstinent. If we do that, then we also have a very
5 clear-cut difference in favor of acamprosate versus
6 placebo. So we do not only rely on these things
7 that make us have some kind of discomfort.
8 We could show it to you. It is in
9 different studies, even. Abstinence per visit is
10 clearly significant in favor of acamprosate as has
11 been shown.
12 DR. G. COOK: Could you comment on how
13 many departures from abstinence might have been
14 missed because of the visit schedule? Do you have
15 a reasonable degree of confidence that the study
16 captured the vast majority of departures of
17 abstinence?
18 DR. MANN: That is, of course, something
19 which I cannot give you exact figures on. This is
20 more what you would call a gut feeling or clinical
21 experience. I think, and you have to be aware of
22 the fact that these patients were not just
23 outpatients which you see maybe three or four or
24 five times. But you have seen them for a week or
25 for two weeks or for three weeks as inpatients and
248
1 you know all about it, and they have already told
2 you how it was and how bad it was and they have
3 already confessed, more or less, that they had all
4 these terrible experiences.
5 Also, their relatives come in. We have
6 talked to their relatives so we know. They don't
7 have anything to hide anymore. If we see them
8 again after six weeks or after twelve weeks, we
9 know that these feelings of guilt and of shame of
10 admitting that you have a relapse, that is
11 something that we have already talked about in the
12 past.
13 If we miss it, then the spouse called us,
14 "How come you don't pick up that he is drinking for
15 the last two weeks?" That is what is happening, or
16 we have this kind of information in 30 to 40
17 percent of our patients throughout the year.
18 So I think we are fairly confident that we
19 picked up most of the relapses during the year and
20 I am very sure that we did not have a difference in
21 picking up those relapses or not between
22 acamprosate or placebo. The same margin of error
23 certainly is true for both groups.
24 DR. OREN: Dr. Hughes
25 DR. HUGHES: I just want to comment on the
249
1 last part that you said which is when we get
2 imprecision, which is the word FDA keeps talking
3 about is precision, you don't worry about it as
4 long too much as long as it is not systematic
5 because what it does is it introduces noise. So
6 what the imprecision does it makes it such that
7 those prior studies had to have a bigger effect in
8 order to detect it.
9 So I almost use the imprecision as an
10 argument that those European trials had a bigger
11 effect and we only found this much of an effect.
12 So, actually, the imprecision doesn't bother me
13 very much.
14 DR. KECK: This is sort of jumping on the
15 same bandwagon, but I think this is the beauty of
16 randomization. It is what randomization should
17 control for especially in a study or studies of a
18 drug that is, from what I can tell--I have never
19 seen anybody in such a trial--virtually
20 indistinguishable from placebo.
21 So the likelihood of unblinding or some
22 kind of systematic, as Dr. Hughes said, bias
23 contributing to the results despite the imprecision
24 of methods I think is pretty small.
25 I guess what I am hung up on a little bit,
250
1 and I would actually appreciate some input from
2 people like Dr. O'Brien and other people who
3 actually done trials in alcoholic patients is Dr.
4 Mason set out a nice table in her slide kit on Page
5 4 comparing the different methods involved in the
6 U.S., which I think is so different than the
7 European studies it is not worth obsessing about
8 anymore, but in the three European studies, how
9 good are these methods because my gut reaction is,
10 in totality, they are not bad.
11 But I want to be comfortable with the .
12 DR. FULLER: You may disagree with me. I
13 don't think they are that bad. Let me try and
14 justify that. It is not uncommon in alcoholism
15 treatment trials, depending on the length of the
16 trial, to interview the person every two or three
17 months. Granted, ideally, you would like to
18 interview them every day, but that is not feasible.
19 Some day, we will have a little wristwatch
20 you can wear that will measure alcohol and we won't
21 be having these discussions. But, until that day
22 arrives, you follow the patient, you track them,
23 you interview them. It is always, then, a
24 retrospective report.
25 Now, the advantage to the time-line
251
1 follow-back is that, hopefully, it improves the
2 accuracy of that or in that patients are given
3 prompts, holidays as indicators of certain days.
4 They are shown these pictures of quantity. So you
5 may get a better frequency, quantity report but,
6 basically, they are both capturing the data, in a
7 sense, retrospectively. The time interval is two
8 to three months.
9 So I think what was done in the European
10 studies was fine. It could have been improved a
11 little bit by current standards.
12 The other comment I will make has to do
13 with randomization. Even if there was somewhat
14 more imprecision in the data collection in the
15 European studies, this should have been randomly
16 distributed across the treatment groups. So I
17 think the data collection is okay.
18 DR. OREN: Dr. O'Brien?
19 DR. O'BRIEN: I really agree with what Dr.
20 Fuller just said. I should tell you all that I
21 have never had any kind of relationship with Lipha,
22 not a consultant or anything like that, but I do go
23 to Europe a lot and I have read all these trials
24 when they first came out and I have heard them
25 presented, both in English and in French. I have
252
1 discussed them when they were fresh.
2 I always was aware of the differences in
3 methodology between the European--as a matter of
4 fact, I have slides of their trials that I have
5 used to compare the kinds of studies we have done
6 here and there. I have used these for years,
7 actually, not just recently, because it has always
8 been very obvious.
9 Then a couple of years ago, I was involved
10 with a group that included Dr. Mann to plan some
11 joint American and European studies of alcoholism
12 using the other medication that has been talked
13 about here, naltrexone, a depo form of it. So I
14 think we had people representing many of the
15 European countries where these studies were done.
16 We arrived at combined protocols. But, in
17 the past, they really were different. But, at the
18 same time, I was always impressed and I still am,
19 that there is an effective drug there and that,
20 while I always had problems with the design of the
21 studies, the way they originally were done, I still
22 felt that there was some efficacy there. That is
23 also borne out by my talking with clinicians in
24 Europe who, in fact, believe, for what it is worth,
25 that the drugs are effective.
253
1 DR. OREN: Dr. Schatzberg?
2 DR. SCHATZBERG: I have a question for the
3 FDA staff. In terms of the PRAMA study, which had
4 longer intervals going out, were you folks
5 satisfied that, in the first 120 days where you had
6 more frequent interviews of the patients, that the
7 drugs separated in terms of either time to first
8 drink, as was presented earlier by Dr. Mann, or in
9 terms of total abstinence because I think if there
10 is an effect still at the 120 days, which is a
11 reasonable length of time for these folks, that
12 would connote substantial benefit for the large
13 group of patients and would still be within that
14 time of frequent assessment so you wouldn't have to
15 worry about whether you are, in fact, having some
16 sort of systematic effect in terms of recall.
17 DR. WINCHELL: I didn't look at 120 days.
18 I know that Dr. Wang replicated the
19 time-to-first-relapse analysis.
20 DR. SCHATZBERG: You did?
21 DR. WANG: As she showed you on her slide,
22 there is a delay of the time to first relapse that
23 comes out statistically significant.
24 DR. SCHATZBERG: Even if you just go to
25 120 days?
254
1 DR. WINCHELL: Oh; I don't.
2 DR. WANG: I didn't specifically look at
3 120 days, either, but what I would like to point
4 out for the PRAMA study is time to first relapse is
5 the prespecified primary efficacy endpoint. This
6 is the only study that prespecified and had a
7 result coming out consistent with other endpoints.
8 What I am really struggling with was there
9 was a question asked from the committee whether the
10 company used the same model to do the European
11 studies. Because I did so many different analyses
12 in trying to understand what is going on, if what
13 we are seeing here from the U.S. trial is true,
14 which means that the acamprosate median dose has a
15 shorter treatment exposure, more dropouts, by that
16 kind of modeling adjustment, it to make the worst
17 outcome to be better.
18 If this logic applies, then the European
19 trials, using the same kind of definition, it
20 should be in favor of placebo, logically.
21 DR. OREN: Sometimes, the wisest people
22 are silent. I know, Dr. Porrino, you haven't said
23 much today. I wonder if you might share some of
24 your thoughts on this efficacy question.
25 DR. PORRINO: Part of my silence really
255
1 comes from the fact that I am a basic scientist who
2 is now starting to dabble in looking at human
3 patients and, in particular, alcoholics. I don't
4 conduct clinical trials, so I consider this a
5 remarkable learning experience for me and I
6 appreciate the opportunity to be a part of this
7 because I have learned a tremendous amount.
8 But one of the things that keeps coming
9 up--there are two things that I could comment on.
10 One of them is the discussion of motivation,
11 motivation as an important variable, and the
12 difference between motivation to completely stop,
13 to remain completely abstinent, and those that are
14 willing to slip a little.
15 In our experience, and this is not just
16 experience with alcoholics where I have much less
17 experience, but with marijuana users where I have a
18 tremendous amount of experience. We have looked at
19 subjects at that point and we have asked them sort
20 of that very question, although not exactly phrased
21 that way, and then we have done some brain
22 imagining.
23 I will say that there is a large
24 difference between the brains of those individuals
25 who are willing to slip occasionally and those that
256
1 are really trying. So motivation is a very
2 important variable and I don't think it should be
3 underestimated nor do I think that combining the
4 two is necessarily appropriate.
5 So I appreciate that it sounds the same
6 and very often is the same, but, actually, in our
7 hands, it looked quite different. Their brains
8 looked quite different so I was quite interested in
9 putting those two together versus separating them
10 which I think is a more appropriate thing to do.
11 The other thing that I can comment on is
12 the fact that, in the patients that I have seen and
13 the alcoholics that I have seen, there is a
14 tremendous desire to have aids and any possible
15 chances to try and remain abstinent. They want to
16 get better, at least many of the ones that I see.
17 And there are no ways to help them.
18 So acamprosate, although it may not be the
19 perfect drug, may certainly work for some where
20 other drugs don't work. I think we need to
21 consider that very importantly.
22 DR. OREN: Dr. Malone?
23 DR. MALONE: I don't really work with
24 drugs and alcohol either, but, in looking at the
25 result of the American study, I think the problem
257
1 that it didn't find any result, I guess, makes us
2 look more closely at the European studies. So it
3 seems that they were using older methodologies and
4 they didn't have preplanning which is troubling.
5 Then I think you start thinking about the
6 way we deliver medical care now and you wonder
7 whether the results from those older studies will
8 be applicable in the way we deliver care in the
9 United States right now for efficacy.
10 DR. OREN: Beyond that, as a child
11 psychiatrist, there is no data presented with
12 regard to alcoholism in youth. Do you have any
13 thoughts on that?
14 DR. MALONE: We study conduct disorder. I
15 guess maybe these children might go on to drink.
16 They might drink now and we don't really know. We
17 have the same problems with following out
18 populations. Half of them never come back to the
19 studies.
20 But I think one of the things that we did
21 learn is that it seems to me that some of the
22 treatments work better in one setting than another.
23 So, for instance, you might have a treatment that
24 works pretty well in an inpatient controlled
25 setting, but when you take it to the outpatient
258
1 setting, it doesn't seem to work as well.
2 So this is really the problem I have with
3 the older European data is that it really is about
4 a treatment for a different setting. The only data
5 we have in the current American setting is negative
6 data. Overall, I think that does cast some doubt
7 on the efficacy of using that dataset to say
8 whether the drug will work the way it is used in
9 the United States, the way it would be used, people
10 not getting detoxed, and maybe being on drugs,
11 polydrugs, when they start the treatment.
12 DR. OREN: Dr. Winokur?
13 DR. WINOKUR: I had wanted to come back to
14 the issues that I had raised before but directed to
15 the FDA representatives, Dr. McCormick or Dr.
16 Winchell, and Dr. Malone came back to that
17 beautifully. So I just wanted to follow up on
18 that.
19 One possibility might have been that we
20 have had data from the U.S. study that supported
21 efficacy and then we could put that together with
22 the European studies that were done a bit ago, but
23 also have some data supporting efficacy and look at
24 them together. As it has happened, we generally
25 agreed that we are going to have to primarily look
259
1 at the European studies and think through how
2 convincing we find the efficacy data to guide our
3 thoughts.
4 We have heard from Dr. McCormick that
5 there is precedent or openness to consider data
6 from the European trials to form an opinion for
7 approval, but, I guess the concern that I had
8 thought about, and Dr. Malone expressed, is if
9 there are differences between the clinical
10 circumstances in the European studies in this case
11 done a while ago and what we have heard to be the
12 case currently in the U.S., and we are talking
13 about a U.S. approval, does that represent a
14 problem from the agency's point of view in terms of
15 that being the exclusive basis in terms of efficacy
16 data?
17 What I am explicitly thinking about is the
18 use of the inpatient detox as a lead-in to having
19 abstinent patients to begin the trial which was
20 done in Europe we have heard is rarely possible in
21 the U.S. We have seen that when a study was
22 launched in the U.S. with the intention of having
23 abstinent patients, there was a very high degree of
24 lack of success in achieving that.
25 So I would like to hear some response from
260
1 the FDA.
2 DR. McCORMICK: I don't believe that that
3 would be a problem. There are ways to abstinence
4 that are nonpharmacologic. So I guess that is
5 another question that we have to you. I guess that
6 is really the essence of the third question, are
7 there subsets that we could identify that might be
8 more responsive and is abstinent prior to
9 initiation of treatment necessary.
10 But the approval of this product, based on
11 European data, given a different set of medical
12 conditions, would not preclude our approval of this
13 product.
14 DR. OREN: Dr. Schatzberg?
15 DR. SCHATZBERG: It would seem to me that
16 the only positive data you have are in abstinent,
17 fully abstinent, detoxified patients so that there
18 are no data that we have seen that it works,
19 particularly in the U.S. trial--that if you are not
20 detoxified, it will have any effect. So I would
21 think that that one group would have to be there
22 because I think it would be misleading to imply
23 that to the public that you could just sort of hand
24 it out in your office to an actively drinking
25 subject and you are going to have any efficacy that
261
1 is true.
2 Just a couple of comments because I am
3 going back to the West Coast. I think the FDA has
4 done a service, in a way, to the sponsor in going
5 that extra mile to look at the European database to
6 see if there is something that can be common across
7 the studies in terms of looking at abstinence and
8 brought some clarity.
9 From a consultant's end, we can't comment
10 on the quality of the data because we don't have
11 the books. We really don't know what they look
12 like, but fact that there is some assurance that
13 two or three of the trials, with the drugs
14 separated on a very highly conservative measure,
15 that does have public-health significance and
16 really ought to count in spite of the fact that you
17 have a failed or a negative U.S. trial where you
18 can't say anything except that it didn't work and
19 there was a high placebo-response rate and a high
20 dropout rate, which are two kisses of death, I
21 think, for clinical trials.
22 But I think you and your staff ought to be
23 given some kudos for really trying to bring clarity
24 on this problem although I am not sure that any of
25 us, either as consultants or people on the
262
1 committee, can tell you what the data looks like.
2 You have got those data right there.
3 DR. McCORMICK: Thank you.
4 DR. OREN: Dr. Ortiz, I know you have been
5 on the left so I haven't always looked straight at
6 you. Is there anything you might want to
7 contribute?
8 DR. ORTIZ: No. I actually had just
9 written down some thoughts. Since we had left
10 Question No. 1, although it seems like we seem to
11 be moving in a direction that the differences can't
12 really be reconciled very well, and we were on
13 Question No. 2, I had come to the same conclusion
14 that Dr. Schatzberg had addressed, that we clearly,
15 I think, seem to have evidence that it is an
16 effective medication for abstinent alcoholic
17 patients.
18 DR. OREN: Dr. Hamer?
19 DR. HAMER: For me, I think the U.S. study
20 is sort of off the table. I think that the
21 decisions need to be based on the European studies.
22 Also, with respect to American study, I want to
23 drag in some really trite, elementary statistics
24 and just remind everyone that failure to reject the
25 null hypothesis doesn't prove the null hypothesis
263
1 is true.
2 So, merely because, in that U.S. study, we
3 failed to show that acamprosate beat placebo
4 doesn't prove that it doesn't beat placebo. All
5 the noise in the world will just make it look
6 worse. That doesn't carry as much weight. I am
7 reassured that the reanalyses that the FDA carried
8 out with some fairly hard endpoints in a
9 conservative way, in a relatively precisely defined
10 group, as Dr. Schatzberg mentioned, seems to
11 indicate that this at least beats placebo in those
12 trials, and, therefore, as an additional weapon in
13 the armamentarium that is fairly sparse right now,
14 might have some use in medical practice.
15 DR. OREN: Dr. Fuller?
16 DR. FULLER: I second those comments. I
17 am persuaded--I think, from the European data, that
18 acamprosate has some efficacy and it is really
19 based somewhat on the literature. Some of these
20 studies were published before. Of course, the
21 problem with the literature, I recognize you don't
22 have the full report, also, by the material that
23 was presented here, and Dr. Winchell's summary of
24 those reports.
25 So I would second the last two comments,
264
1 that the European data do indicate efficacy.
2 DR. OREN: Dr. Mehta. Then we are going
3 to one-by-one through everyone to ask you to
4 register your opinion.
5 DR. MEHTA: Just a comment to what Dr.
6 Malone said. Dr. Goodman showed a slide which
7 showed that the core illness for alcohol dependence
8 is similar in the U.S. and in Europe. This was
9 shown based on a letter written to FDA by NIAAA.
10 DR. OREN: Dr. Malone; you have a
11 question?
12 DR. MALONE: No; the study populations
13 were very different, though, because the European
14 one did not really include people who had abusive
15 drugs and it didn't include people who were
16 drinking. So even if just alcoholism is the same,
17 the study populations were very different.
18 DR. OREN: We have a little more time for
19 commentary, it turns out. Dr. Hughes
20 DR. HUGHES: You know, the thing that is
21 hanging me up, and let me try to put it as an
22 analogy. It seems to me the analogy is it is like
23 Lipha is a guy who--let's say a baseball player and
24 he has hit a home run thirteen times in a row, and
25 he comes to somebody else and he says, "I can hit a
265
1 home run." And the other person says, "Well, I
2 don't know about that."
3 And the guy says, "Well, I tell you what.
4 I will prove it to you. I will do it right now."
5 And he tries to do it right now and he doesn't hit
6 the home run. We know he has hit it thirteen times
7 in a row but he put himself at risk by saying, he
8 can prove it to you to you that next time.
9 So what I am hung up on is, as a result of
10 this trial, I am less confident that this drug
11 works than I was at the get-go. So I am a little
12 bit worried about the precedent. In other words,
13 what would have Lipha had to have done in this
14 trial to disprove it. I am not sure what they
15 would have had to have done for us to say, "You
16 can't have approval."
17 Then, as a result, I worry about the
18 precedent there; that is, that it seems to me that
19 if you make an agreement, that you agree that you
20 have to show your drug works in a subset before you
21 are going to get approval and then you don't get
22 it, that is what we used to call going back in your
23 word.
24 So that is where I am hung up.
25 DR. OREN: Or, to use your baseball
266
1 analogy, perhaps when a ball player was younger in
2 the different town, he could hit home runs. But it
3 a few years later and he is in a different city and
4 time has passed a little.
5 Dr. O'Brien, did you want to say
6 something?
7 DR. O'BRIEN: Before the baseball, we were
8 talking about detoxified patients. I just wanted
9 to point out that, while it is the mode right now
10 to not admit people for detox or even pay that much
11 for outpatient care, if, indeed, there were
12 evidence about the state of a patient--in other
13 words, if this is emphasized that the people should
14 be drug free before they start on the medication,
15 then this probably would be cost-effective--in
16 other words, to invest something in a
17 detoxification, to start them off clean--because
18 what you would pay at the outset, even if you had
19 to admit them for a few days would be more than
20 offset, if you were an HMO, by the savings over the
21 next few years.
22 We already heard that Kaiser Permanente is
23 using another drug which is reasonably expensive
24 and they must be doing it because it is
25 cost-effective. I think there are data showing
267
1 that it is cost-effective.
2 So I think that we needn't worry about the
3 fact that, in the American trial, there weren't a
4 lot of people who were abstinent at the beginning
5 because there could have been if, in fact, that had
6 been a requirement.
7 DR. HAMER: I just want to continue the
8 baseball analogy a little bit. I think what has
9 happened here might be that the baseball player hit
10 thirteen home runs and then made the wager with his
11 friend. Then, after they agreed, the friend said,
12 "Oh; by the way, for this at bat, we are using a
13 smaller baseball, you are getting a lighter bat and
14 the pitcher is a foot and a half taller and has
15 been lifting weights for the last five years."
16 DR. OREN: Dr. Malone?
17 DR. MALONE: Back to what Dr. Hughes said,
18 the problem was that the American trial was
19 negative. Was the American trial necessary? They
20 could not have come forward with just the European
21 trial? I don't quite understand.
22 DR. OREN: Do you want to repeat the
23 question?
24 DR. MALONE: Back to what Dr. Hughes was
25 saying. You have these positive trials and now
268
1 you, somehow, come here to the FDA and you do
2 another trial and it is negative. I would think
3 that would put you in a worse position unless that
4 trial was somehow not necessary.
5 DR. McCORMICK: I guess, to go back to the
6 baseball analogy, we don't expect all home runs.
7 As I mentioned this morning, we frequently do see
8 development programs in which there are trials
9 which may trend in the right direction but are not
10 statistically significant on the primary endpoints
11 and, occasionally, we see some that really show no
12 effect at all.
13 We try to understand why that is the case.
14 We try to assure ourselves, as we are in this case,
15 that the studies that we are relying upon, or the
16 studies that are positive, aren't fallacious.
17 First of all, let me just set the record
18 straight. There aren't thirteen home runs. Let's
19 just say the three pivotal studies that we have
20 reviewed may be characterized as home runs. I see
21 a difference of opinion which we would like to hear
22 from, but the fact that there is a negative study
23 doesn't trouble us. It is not a preclusion to
24 approval.
25 DR. OREN: Dr. Malone?
269
1 DR. MALONE: Was the purpose of the
2 American study for efficacy or really just safety
3 in the different sample that you get in the United
4 States?
5 DR. WINCHELL: The purpose of the American
6 study, as we understood it when we first met with
7 the company, was because they wanted to make a
8 change from marketing the 333-milligram tablet to
9 the 500-milligram tablet. So, since there were no
10 studies on the 500-milligram tablet, what we agreed
11 to do was accept a marketing application that
12 consisted of a single study using the 500-milligram
13 tablet with a nominally very similar total daily
14 dose of 2 grams, although we didn't expect that
15 complete bioequivalence, as we define it, would be
16 established.
17 We said, okay; if you can do one winning
18 study with the 500-milligram tablet, the other
19 stuff you have got here on the 333 milligrams, two
20 tablets TID, will serve as your supportive evidence
21 of efficacy, your confirmatory evidence. That is
22 how this whole story began.
23 DR. McCORMICK: If I can just add another
24 word. We did conceive of this as an efficacy study
25 and a safety study and it was designed to obtain
270
1 efficacy information and proof of efficacy.
2 DR. OREN: Dr. Cook
3 DR. COOK: I want to refer to the thirteen
4 home runs again. First of all, there are three
5 studies submitted besides the U.S. study, so there
6 can only be three home runs. Number two, some that
7 are not submitted were not positive, at least one.
8 Number three, I count three studies. Based on the
9 analyses, number one, you could consider none of
10 them at bats on the basis of no
11 prospective-analysis plan.
12 So, to go beyond that is to bend over
13 backwards, I think. I don't care if it was 1988,
14 if we were in the clinical-research center at any
15 major university, if you didn't have a prospective
16 data-analysis plan, the study wouldn't go through.
17 This study would not have been approved for funding
18 at most institutions.
19 Then, if we look at the analysis, two
20 studies seem to be positive, the Pelc II and the
21 Paille. Dr. Wang I think was fairly convincing
22 that, unless you look at it just the right way, the
23 Paille was not. Again, you have to be conservative
24 if you didn't prespecify the analysis.
25 Now we have two studies. That is enough
271
1 in the analogy that two hits out of four is a
2 pretty good batting average or the idea that more
3 than two well-conducted studies have been positive
4 Now, I have already said I have a problem
5 with well-conducted. But, seeing that this hasn't
6 been monitored, anything in the monitoring that
7 doesn't show that randomization was perfect, that
8 everything was on the up-and-up, in me, may be
9 based on what we have that is tentative and not
10 fully monitored. But, it is very slippery.
11 Anything that is weaker than it already is in those
12 studies is a problem.
13 I worry about the differential dropout
14 rate with placebo in those studies. That is why I
15 am concerned about randomization.
16 DR. OREN: I am going to try and move on a
17 little bit. Before we go on a person-by-person
18 vote, I just wanted to ask the members of the
19 committee if any of you wanted to make any general
20 statement before we each register our opinions.
21 What I will do is I will go
22 person-by-person asking you to say yes, no, or
23 abstain. If you wish, you can argue at that point
24 or share some of your rationale for your vote if
25 you would like. But, before we register those,
272
1 does anybody want to make any additional point from
2 the committee or from the guests?
3 What I would like to do is, for the
4 nonvoting members of the committee, I just want you
5 to say if you were voting, please share with us how
6 you might vote and why you might do that, although
7 you are obviously not voting.
8 Dr. Mehta?
9 DR. MEHTA: I just wanted to make a
10 comment that I don't know why we are hung up about
11 the prospective plan for analysis. These studies
12 were done 1998 in Europe. That was the state of
13 the art. Probably these are designed a couple of
14 years earlier. If I go back and look at my own
15 studies in this country and major pharmaceutical
16 companies submitting across all the divisions,
17 these are not very different than what they have
18 done.
19 Maybe in clinical research centers, it
20 would be different. Maybe at different places, it
21 might be different, but certainly not in drug
22 trials, particularly submissions. I have never had
23 any comments from FDA statisticians which says
24 that, look, this protocol or analysis is not
25 acceptable. No. That is absolutely not true.
273
1 DR. OREN: So if you were going to be
2 voting, how would you vote, the question being, is
3 there sufficient evidence of the efficacy of
4 acamprosate in the treatment of alcoholism to
5 warrant approval.
6 DR. MEHTA: Just one additional comment.
7 In another division, the Cardiorenal Division,
8 there was a major ace inhibitor approved for heart
9 failure. The only major and important study was in
10 the United States. It was totally negative. Bob
11 Temple said they had tried, just like what you have
12 done, about twenty different ways of looking at the
13 data to find out if there was some redeeming
14 feature in that study. There was none.
15 Nevertheless, based on the two or three
16 European studies, the drug was approved and it is
17 on the market. Subsequently, several years later,
18 there was an American positive study.
19 All right. Coming back to this drug, I
20 would approve it because there are three studies
21 which have been shown that the drug is clearly
22 different than placebo. The U.S. study, I would
23 just ignore it. It is three to one, batting
24 average.
25 DR. OREN: Thank you.
274
1 Dr. Hughes, if you were voting, what would
2 you tell us?
3 DR. HUGHES: Vote for approval.
4 DR. OREN: Any additional comment? No?
5 Dr. Porrino?
6 DR. PORRINO: I vote for approval.
7 DR. OREN: Dr. O'Brien? You are obviously
8 influential in the field of alcoholism and whatever
9 you think will clearly have a great impact. So,
10 although you are not voting, tell us how you would.
11 DR. O'BRIEN: Well, first of all, I would
12 like to say that I was extremely impressed with the
13 material that the FDA gave us to prepare for this.
14 I was already familiar with most of these papers.
15 I had reviewed some of them for publication. This
16 was the best exposition I had seen. Drs. Winchell
17 and Wang gave just beautiful presentations this
18 morning.
19 I think they were correctly very rigorous.
20 So certainly I will have to say that, if I had been
21 asked this question before I got these materials
22 and heard them, I would have been much more
23 positive about the drug. But I still feel, and it
24 is hard for me to separate the three studies from
25 what I know about the other group of studies, I
275
1 would consider two of the other studies not to be
2 positive and all the rest of them, for various
3 reasons, I don't have to go into here--but, in
4 other words, the vast majority were positive.
5 To me, it is remarkable that they were
6 positive because of the imprecision involved, I am
7 critical of some of the design, and also because of
8 all of the problems with studying this. When we
9 have situations where, with antidepressants, there
10 is evidence that 50 percent of the trials fail to
11 show an advantage for a so-called active drug over
12 placebo. We had a debate on this at ACNP a couple
13 of years ago.
14 So, anyway, the fact that you could get
15 this much positive with alcoholism must mean that
16 there is efficacy there. So, based on the evidence
17 that we have, if I had a vote, I would have voted
18 positive.
19 DR. OREN: Thank you.
20 Dr. Fuller, you do have a vote, so please
21 tell us.
22 DR. FULLER: I am going to make this five
23 or six hits in a row. I find, I think as I
24 expressed earlier, the European data are reasonably
25 credible. I think the method of collection of data
276
1 was reasonably standard. I believe in many of the
2 studies they did breath alcohols at the time of the
3 interview and these are little tricks that are done
4 to try and improve the quality of data.
5 The differential dropout rate in the
6 European studies actually I think is in favor of
7 the medication. My thinking is along these lines.
8 I think the placebo patients felt that they weren't
9 getting something out of the treatment so they were
10 more likely to drop out of treatment.
11 Now, one can always think of caveats.
12 Certainly, if there were problems
13 post-randomization that are not apparent from the
14 material that was given, that would influence me.
15 But, taking it as a whole, the material that was
16 given with its pros and cons, with the summaries
17 prepared by Drs. Winchell and Wang, and based sort
18 of on my clinical and other research experience,
19 weighing all these, I think the European data
20 indicates there is some efficacy for acamprosate
21 and it should be approved.
22 DR. OREN: Dr. Cook
23 DR. COOK: I have one general comment that
24 I can't leave without stating. I don't want to
25 minimize the importance of motivation in treatment
277
1 but I don't want patients who participated in
2 trials as described as less than motivated because
3 my view is, whether people are abstinent or not,
4 they are motivated to stop this.
5 I particularly want to point out whether
6 people's goal was different. The issue is how they
7 answered the question. The question was, I seek
8 total abstinence versus I seek total abstinence but
9 realize I may slip. I realize I may not be
10 perfect. That actually may be a step in the right
11 direction to somebody who is recognizing they don't
12 have complete control over themselves.
13 Had the question been, my goal is complete
14 abstinence, or my goal is complete abstinence with
15 a few slips, that is a different question. So, I
16 have struggled with this, obviously a lot, and I
17 guess I said before, I do see two positive studies,
18 Pelc and PRAMA, no matter how it is looked at and
19 the only question is verification.
20 So I guess I say yes with that caveat.
21 DR. OREN: Dr. Ortiz?
22 DR. ORTIZ: I am very appreciative of the
23 FDA staff for bringing this confusing picture to us
24 from around the country and to the public to
25 consider what to recommend for the American public
278
1 given some of this confounding data and confusing
2 data.
3 I was very confused at home going over the
4 data. But I also realize, again having the
5 gentleman from Kaiser that represents, basically,
6 the working alcoholic in the United States that is
7 insured and their willingness to use new
8 medications for this group, in thinking about my
9 population from New Mexico which is a rural
10 population with lots of Hispanics and Native
11 Americans, I guess, again, again going back to the
12 American study, I am concerned that it doesn't
13 represent what the American alcoholic is like.
14 It seems that the issue is really what is
15 shown by the European studies and I also concur
16 that they do appear to show efficacy.
17 DR. OREN: Let's go down to the other side
18 of the table. Dr. Leon?
19 DR. LEON: I have expressed my concerns
20 about the methodology, the prospective--I mean,
21 getting to the home-run analogy, I feel like the
22 fence was moved after the ball landed, as you have
23 heard me say that many times today.
24 So I vote against it. I think there is a
25 need for another study with more rigorous,
279
1 prospectively defined--that is, defined before the
2 first subject is enrolled--more rigorous
3 methodology using the assessment procedures of the
4 U.S. study.
5 DR. OREN: Dr. Keck?
6 DR. KECK: I am not going to use the
7 baseball analogy. I am actually going to limit my
8 remarks because they have already been well
9 expressed by Drs. Fuller and O'Brien. I will vote
10 in the affirmative.
11 DR. OREN: Dr. Hamer?
12 DR. HAMER: I hate to disagree slightly
13 with my colleague Dr. Leon, but in terms of the
14 prespecified endpoint, I am reminded of an incident
15 four or five or six years ago in cardiorenal in
16 which a clinical trial was stopped early because so
17 many fewer patients were dying with placebo than
18 with drug and then the sponsor had a great deal of
19 trouble getting it approved because death was not a
20 prespecified endpoint.
21 We need to be rigorous, but I think we
22 need to put a great deal of thought into it. I
23 especially complemented the FDA reviewers earlier
24 in person and I want to complement them publicly on
25 the absolutely thorough coherent job they did with
280
1 this material. My vote would be in favor of
2 efficacy.
3 DR. OREN: Dr. Winokur?
4 DR. WINOKUR: I also vote in favor of
5 efficacy based on the European studies. I echo all
6 the comments about the extremely high quality of
7 their review and presentation by the FDA reviewers.
8 I guess my other comment is, even though
9 I, and many of us, have stated the opinion that the
10 data available do meet our standards for
11 demonstration of efficacy, it is also clear, and
12 especially in the discussion of the U.S. trial,
13 that there is an awful lot more to be learned. I
14 would hope that the sponsor and the investigators
15 in the field would continue to work forward to
16 understand more about the complex variables that
17 are related to effective use of this agent.
18 DR. OREN: Dr. Malone?
19 DR. MALONE: I think everything taken
20 together, I would say that it seems to be
21 efficacious in the sample who undergo detox and are
22 abstinent at the time of starting the drug. But I
23 think, for other samples, you don't have any data
24 for efficacy. So, for that one sample. And the
25 American sample might really end up being different
281
1 because maybe the alcohols in the United States
2 tend to use what seemed, from the data, a lot of
3 drugs and they are not going to be abstinent when
4 they start taking the medicine.
5 DR. OREN: Actually, we will come to
6 samples as a part of our last question.
7 Dr. Rudorfer?
8 DR. RUDORFER: I would like echo what Dr.
9 Malone just said. I am troubled by the American
10 study in that it seems to have been the best
11 conducted one and I think Dr. McCormick used the
12 term this morning about the targets of the drug,
13 were it to be approved and the U.S. study actually
14 consisted of the real targets.
15 Having said that, I am persuaded that at
16 least two of the European studies did show efficacy
17 under narrowly defined conditions. Patients who
18 were medically detoxified, and even if it is hard
19 to do inpatient nowadays in the U.S., it can be
20 done on an outpatient basis and people who were
21 abstinent on entry to the study, I believe did
22 benefit from the drug. So, overall, I would vote
23 in the affirmative.
24 DR. OREN: For my vote, just so Dr. Leon
25 won't be alone, I will join you in voting in the
282
1 negative although that is the minority vote. I
2 think that it is not unreasonable to hold a drug to
3 current standards even if the data are from the
4 past. About fifteen years ago, I bought a
5 townhouse from a chronic alcoholic who was one of
6 the designers of the Challenger space shuttle that
7 crashed. If we were trying to evaluate a new
8 proposal for a space-shuttle design and we were
9 being submitted with the original standards because
10 they were good enough in that time, I am sure that
11 we would not accept that because we have learned
12 something since then.
13 I think it behooves us to try and take the
14 latest knowledge and use it and make the best
15 possible use of it. So, although the narrow
16 circumstances of the European studies, I hear them,
17 I am not fully persuaded by them.
18 Having said that, if the FDA
19 were--clearly, there is a strong sense of a
20 majority opinion to encourage the FDA to approve
21 the drug, my encouragement, and this was be the
22 segue into the last question for us to talk about
23 which is do the data support any conclusions
24 regarding subgroups, I didn't hear the sponsors
25 describe the drug as being a home-run hitter.
283
1 It wasn't described as a panacea. It
2 wasn't a lithium, a penicillin, a fluoxetine. So I
3 think it would be very important that, if the drug
4 were to be approved, that the indications for it be
5 very clearly identified and we should talk about
6 what those indications might be.
7 I would encourage, certainly, the FDA to
8 not be reticent about describing those indications
9 and not hesitate about the marketing of the drug,
10 that its limited value be not overstated in the
11 marketing.
12 So maybe this would be a good time to turn
13 then to the last question which is do the data
14 support any conclusions regarding subgroups and
15 this might give the FDA some guidance in--
16 MS. TITUS: I just want to do a formal
17 vote into the record so there are no phone calls
18 back to me later on what the formal vote was. It
19 was eight yesses, two nos and, of the eight yesses,
20 there were several conditions attached to that
21 which you will see in the transcript when it comes
22 through.
23 DR. OREN: Okay. On the last question,
24 does anybody want to offer some comments or
25 suggested answers
284
1 Dr. Hughes?
2 DR. HUGHES: I think it would be very
3 important that the FDA replicate the analyses on
4 the 4500. I thought the way the FDA went through
5 the different hypotheses of subgroups, is it
6 severity, is it behavior therapy, is it motivation,
7 is it abstinence, et cetera, that if we did that
8 same sort of analysis with this larger sample size,
9 that would be a very good way to decide on any
10 subgroups.
11 DR. WINCHELL: We would need that efficacy
12 data. We do have the integrated safety data but I
13 don't believe we have got the efficacy data.
14 DR. McCORMICK: We do have the efficacy
15 data on the three European studies that we have
16 been discussing, so that would be feasible.
17 DR. HUGHES: I guess, since I am not a
18 member of anything, I would really encourage Lipha
19 to provide the data of the 4500 patients so that
20 you can replicate that or perhaps some third
21 disinterested party could replicate that, I think
22 would be very important because I think that is
23 your best data source for deciding whether or not
24 to restrict the use to a subgroup.
25 DR. OREN: Dr. Leon?
285
1 DR. LEON: A point of clarification. I
2 know in one of these documents, it not only
3 mentioned that the indication was for the
4 maintenance of abstinence but also for they
5 recommended one year of treatment. Is that part of
6 this vote, or part of this discussion? It is?
7 Okay. I just want to point out, in my looking at
8 the data which I did, I notice that actually none
9 of the trials treated anyone for a full year. One
10 of them came close, 48 weeks.
11 That was the PRAMA trial. In that, only
12 79 subjects out of the subjects who were enrolled,
13 on active medication completed the trial. So I
14 don't think there is a lot of data there supporting
15 one year of treatment.
16 There is actually no data there supporting
17 one year of treatment and there are 79 subjects
18 that went 48 weeks.
19 DR. HUGHES: If I could comment on that.
20 It is often with medications, physicians use longer
21 durations than are labeled, so, especially with
22 drug-dependent patients in which oftentimes many
23 clinicians feel like a longer duration is
24 warranted, I would hope there would be some
25 flexibility around that duration because I know, in
286
1 my field, I have done a lot.
2 There was, early on, a statement that you
3 should not use agonist therapy beyond a certain
4 point, should not do this. I think that has been
5 somewhat harmful to field. I would rather see use
6 beyond some point at the discretion of the
7 prescribing physician.
8 DR. OREN: Dr. Rudorfer?
9 DR. RUDORFER: Just another comment and
10 then maybe a question to the FDA related to that.
11 We are specifically not addressing safety issues at
12 this meeting but, in real life, if the drug were
13 approved, of course, physicians would need to
14 consider the benefit-to-risk ratio which I would
15 assume that issues like duration of treatment
16 should be considered at that time.
17 So, for instance, if there are adverse
18 effects that only appear after six or eight or ten
19 months, then that may well influence the length of
20 treatment.
21 DR. McCORMICK: You are absolutely right.
22 We are looking at that and will have that
23 information within the next few weeks.
24 DR. OREN: Two of the predictors, or
25 positive predictors, of good response from the drug
287
1 were someone being detoxified before starting the
2 use of it and being committed to abstinence. Does
3 the committee accept these particular subgroups and
4 should this be something that the FDA should,
5 perhaps, encourage in its labeling or in terms of
6 marketing or indications?
7 Dr. O'Brien?
8 DR. O'BRIEN: The one about abstinence is
9 something which is physiological. You can think of
10 a lot of other situations in which a recommendation
11 about the use of a drug is dependent upon a
12 particular state that someone is in. So I think it
13 is pretty clear-cut and you can even verify it with
14 the appropriate tests.
15 The one about the motivation is much more
16 difficult because, with all due respect to the
17 questionnaires that were used, no one would really
18 expect that an alcoholic or any other person who
19 has been diagnosed with a substance-use disorder
20 has any consistent level of motivation.
21 We actually have motivational scales that
22 we use that would get at it more specifically, but
23 ambivalence is one of the hallmarks of this
24 disorder so that a person may tell you one minute
25 that, I am totally motivated to be abstinent for
288
1 the rest of my life and walks out of your office
2 and starts drinking again.
3 This happens all the time. It is not that
4 they were lying in one case. It is just that they
5 are impulsive and things change. So I am not so
6 sure that we would gain very much by that, but I am
7 in favor of recommending that people not use the
8 drug until they achieve abstinence and then it is a
9 drug for maintaining abstinence rather than helping
10 to induce abstinence.
11 DR. OREN: Dr. Malone?
12 DR. MALONE: It seemed also from that data
13 that they would have to be abstinent from other
14 substances, so it wouldn't just be alcohol. You
15 shouldn't be abusing other substances, it seemed to
16 me, at least, comparing the American and European
17 data, that was one of the key differences, was
18 using other substances.
19 DR. OREN: Dr. Winokur?
20 DR. WINOKUR: Just to reinforce that, I
21 think it is important to point out that the only
22 data that we had a chance to look at where we did
23 see efficacy was under circumstances where
24 abstinence was the case at the time of instituting
25 treatment, and the study that didn't go that way,
289
1 there was a more complicated situation.
2 So, until we have other data to broaden
3 our understanding, that really has to be the
4 starting point.
5 DR. OREN: Dr. O'Brien?
6 DR. O'BRIEN: I think it has been
7 mentioned but it might be worth highlighting that I
8 believe that one of the studies that most people
9 would--that was negative in Europe was the U.K.
10 study where there was a lot of nonabstinence when
11 they started on the medication. So, in a sense,
12 that certainly supports the conclusion that might
13 draw from the American study and it suggests sort
14 of two-for-two, when they were not abstinent, the
15 results were not better than placebo.
16 DR. OREN: Any additional comments from
17 the committee? Do the FDA staff want us to address
18 any other particular aspects?
19 DR. McCORMICK: No. I would like to thank
20 you. This discussion this afternoon has been
21 extremely helpful for us. You have answered,
22 really, all the questions that we have had. Thank
23 you.
24 DR. OREN: I would like to thank the
25 public who has been here for us, the sponsor for
290
1 presenting their data and, of course, all of
2 members of the committee for your time. I will
3 call this meeting to adjournment. Thank you.
4 [Whereupon, 4:00 p.m., the meeting was
5 adjourned.]
6 - - -