1
DEPARTMENT OF HEALTH AND HUMAN
SERVICES
FOOD AND DRUG
ADMINISTRATION
CENTER FOR DRUG EVALUATION AND
RESEARCH
CARDIOVASCULAR AND RENAL DRUGS
ADVISORY COMMITTEE
Holiday Inn
2
P A R T I C I P A N T S
Dornette Spell-LeSane, M.H.A., NP-C,
Executive
Secretary
Jeffrey S. Borer, M.D., Acting Chairman
Steve Nissen, M.D.
Alan T. Hirsch, M.D.
Thomas Fleming, Ph.D.
Maria H. Sjogren, M.D.
Jonathan
Sackner-Bernstein, M.D.
John R. Teerlink M.D.
Susanna L. Cunningham, Ph.D.
William R. Hiatt, M.D.
Beverly H. Lorell, M.D.
Thomas Pickering, M.D.
Ronald Portman, M.D.
Paul Watkins, M.D., Ph.D.
Jose Vega, M.D.
FDA
Participants
Dr.
Mark Avigan
Dr.
Florence Houn
Dr.
Joyce Korvick
Dr.
Kathy Robie-Suh
3
C O N T E N T S
AGENDA ITEM
PAGE
Call to Order and Introductions - Jeffrey
S. Borer,
M.D., Acting Chair 5
Conflict of Interest Statement - Dornette
Spell-LeSane, NP-C, Executive
Secretary 8
Welcome and Comments - Norman Stockbridge,
M.D.,
Acting Director, Division of
Cardiovascular and
Renal Drug Products, FDA 13
Sponsor Presentation
Introduction - Hamish Cameron, M.D., Vice
President, Exanta 15
Clinical Pharmacology - Troy Sarich,
Ph.D.,
Director, Clinical Pharmacology 24
Efficacy - Jay Horrow, M.D., Senior
Director,
Clinical Development 43
Safety - Sunita Sheth, M.D., Senior
Director,
Clinical Development 97
Benefit and Risk Anticoagulation -
Jonathan L.
Halperin, M.D.,
FDA Presentation
Risk/Benefit Assessment - Ruyi He, M.D.,
Medical
Officer, Division of Gastrointestinal and
Coagulation Drug Products 172
Risk Management of Hepatotoxic Drugs -
Kate
Gelperin, M.D., M.P.H., Medical
Epidemiologist,
Division of Drug Risk Evaluation 282
Drug-Induced Liver Toxicity - Paul
Watkins, M.D.,
Verne S. Caviness Distinguished Professor
of
Medicine, Director, General Clinical
Research
Center,
Medical Center 301
4
C O N T E N T S
(Continued)
AGENDA ITEM
PAGE
Questions from the Committee --
Open Public Hearing 184
Charge to the Committee - Joyce Korvick,
M.D.,
M.P.H., Acting Division Director,
Division of
Gastrointestinal and Coagulation Drug
Products, FDA 318
Committee Discussion 319
Break --
Committee Questions/Summary 320
Adjournment
414
5
P R O C E E D I N G
S
DR.
BORER: It's
call the meeting to order. This is the Cardiovascular and
Renal Drugs Advisory Committee
meeting, and we will discuss New Drug
Application
(NDA) 21-686, proposed trade name Exanta
(ximelagatran) by AstraZeneca, for the
proposed
indication of the prevention of venous
thromboembolism in
patients undergoing knee replacement
surgery, the prevention of stroke and other
thromboembolic complications associated
with atrial
fibrillation, and the long-term secondary
prevention of venous thromboembolic
events after
standard treatment following an episode
of acute
venous thromboembolic event.
We'll begin by introducing
everybody at
the table. In this meeting, the Cardio/Renal
Committee actually is advising the GI
Division as
well as Cardio/Renal, in fact, primarily
the GI
Division, so we have more people at the
table than
we sometimes do. Maybe we can each say our name
and what we're doing here, and we'll
start with Dr.
6
Vega on the far side.
DR. VEGA:
I'm Jose Vega. I'm the
industry representative on the committee,
and I'm
from Amgen.
DR. PICKERING: Tom Pickering from
DR. PORTMAN: Ron Portman from the
DR. HIATT: Bill Hiatt, University of
DR. LORELL: Bev Lorell, Harvard Medical
School, and also Guidant Corporation.
DR. SACKNER-BERNSTEIN: Jonathan
Sackner-Bernstein,
in
DR.
CUNNINGHAM: Susanna Cunningham. I am
the consumer representative
on the committee, and
I'm from the
DR. NISSEN: I'm Steve Nissen. I'm a
cardiologist at the Cleveland Clinic.
DR. WATKINS: Paul Watkins.
I'm a
hepatology consultant from
7
Carolina-Chapel Hill.
DR. BORER: Jeff Borer, cardiologist,
MS.
SPELL-LeSANE: Dornette Spell-LeSane,
Executive Secretary for the
committee.
DR. TEERLINK: John Teerlink, University
of California-San Francisco, and
DR. FLEMING: Tom Fleming, University of
DR. HIRSCH: Alan Hirsch, cardiologist and
vascular medicine specialist at the
University of
DR. AVIGAN: Mark Avigan, Office of Drug
Safety at the FDA.
DR. STOCKBRIDGE: I'm Norman
Stockbridge,
the Acting Director of the Division of
Cardio/Renal
Drug Products at FDA.
DR. HOUN: I'm Florence Houn. I'm the
Office Director for Drug Evaluation III.
DR. KORVICK: Joyce Korvick, Acting
Director, Division of Gastrointestinal
Coagulation
8
Drug Products.
DR. ROBIE-SUH: Kathy Robie-Suh, Acting
Deputy Director, Division of Gastrointestinal
and
Coagulation Drug Products.
DR. BORER: Thank you very much.
We have many people at the
table. I'm
going to remind everyone that when you
speak, you
should press the button on your microphone,
and
when you're done, turn it off, please,
unless you
want to say something because that's the
only way
I'm going to know that you want to if you
press the
button and I see the light.
We'll go on to the conflict of
interest
statement. Dornette Spell-LeSane, the Executive
Secretary of the Cardio/Renal Drug
Advisory
Committee, will present the conflict of
interest.
MS. SPELL-LeSANE: Good morning.
The
following announcement addresses the
issue of
conflict of interest and is made part of
the record
to preclude even the appearance of such
at this
meeting.
Based on the submitted agenda and all
financial interests reported by the committee
9
participants, it has been determined that
all
interests in firms regulated by the
Center for Drug
Evaluation and Research present no
potential for an
appearance of a conflict of interest at
this
meeting, with the following exceptions:
In accordance with 18 U.S.C.
Section
208(b)(3), full waivers have been granted
to the
following participants. Please note that all of
the consulting and speaking activities
waived are
unrelated to Exanta and its competing
products:
Dr. William Hiatt for
consulting for two
competitors for which he receives less
than $10,001
per year per firm;
Dr. Thomas Pickering for
serving on a
competitor's advisory board for which he
receives
less than $10,001 per year;
Dr. Ronald Portman for
consulting for a
competitor for which he receives less
than $10,001
per year;
Dr. Thomas Fleming for
consulting for four
competitors, he receives less than
$10,001 per year
per firm;
10
Dr. Sackner-Bernstein for
consulting for
the sponsor and a competing firm, he
receives less
than $10,001 per year per firm. Also, for his
Speaker Bureau activities for a
competitor, he
receives less than $10,001 to $50,000 per
year;
Dr. Jeffrey Borer for serving on a
steering committee for a competitor, he
receives
less than $10,001 per year;
Dr. Alan Hirsch for lecturing
for the
sponsor, for which he receives less than
$5,001 per
year.
For lecturing for three competing firms, he
receives less than $5,001 per year for
serving on
two Speaker Bureaus, and from $5,001 to
$10,001 for
one Speakers Bureau. Two consulting agreements for
two competing firms, he receives less
than $10,001
per year for one consulting, and from
$10,001 to
$50,000 per year for the other.
In accordance with 18 U.S.C.
208(b)(3), a
limited waiver has been granted to Dr.
Paul Watkins
for serving on two advisory boards for a
competing
firm.
He receives less than $10,001 per year for
one and greater than $50,000 per year for
the
11
other.
Under the terms of this limited waiver, Dr.
Watkins will be permitted to participate
in the
committee's discussion of Exanta. He is, however,
excluded from voting.
Lastly, in accordance with 18
U.S.C.
Section 208(b)(1), full waivers have been
granted
to the following participants for
interests
unrelated to Exanta and its competing
products:
Dr. John Teerlink for speaking
for two
competitors, he receives less than
$10,001 per year
from one, and from $10,001 to $50,000 per
year from
the other. Also, for his consulting for a
competitor for which he receives between
$10,000 to
$50,000 per year;
Dr. Maria Sjogren for
consulting for a
competitor for which she receives less
than $10,001
per year.
A copy of the waiver statement
may be
obtained by submitting a written request
to the
agency's Freedom of Information Office,
Room 12A-30
of the
discussions involve any other products or
firms not
12
already on the agenda for which FDA
participants
have a financial interest, the participants
are
aware of the need to exclude themselves
from such
involvement, and their exclusion will be
noted for
the record.
We would also like to note that
Dr. Jose
Vega has been invited to participate as
an industry
representative acting on behalf of
regulated
industry.
Dr. Vega is employed by Amgen.
With respect to all other
participants, we
ask in the interest of fairness that they
address
any current or previous financial
involvement with
any firm whose products they may wish to
comment
upon.
Thank you.
DR. BORER: Thank you very much, Dornette.
That was about the longest conflict of
interest
statement that I can remember.
But we still are five minutes
ahead,
Norman Stockbridge, the Acting Director
of the
Division of Cardiovascular and Renal Drug
Products.
13
DR. STOCKBRIDGE: I'll see if I can keep
us on schedule. Good morning and welcome to what
promises to be an interesting meeting on
behalf of
the Divisions of Cardio/Renal Drug
Products and GI
and Coagulation Drug Products. I want to thank
members of the Cardio/Renal Advisory
Committee,
consultants, and the sponsor for their
participation.
I do need to acknowledge retirement
of
four members from the Advisory
Committee: Alan
Hirsch is here today, a couple of chairs
down to my
left; Steve Nissen is over at the middle
of the
table there; Paul Armstrong would be here
today
except that Homeland Security discovered
that he's
a Canadian.
[Laughter.]
DR. STOCKBRIDGE: And, finally, there is
our Chairman, Dr. Jeff Borer. Dr. Borer's service
to the committee began in 1977, an era in
which
members still sported powdered wigs.
[Laughter.]
DR. STOCKBRIDGE: I can't quite tell from
14
the records where he cast his first vote,
but in
that year, the committee heard arguments
on
potassium and atropine.
As tokens of our appreciation,
Ms.
Spell-LeSane has for each of you some
actual
certificates signed by our Acting
Commissioner and
some virtual plaques that look just like
this one.
So on behalf of Cardio/Renal, the Food
and Drug
Administration, and a grateful nation,
thanks to
you all.
[Applause.]
DR. NISSEN: Norman, I'm not from Canada,
but I'm from Cleveland, and it's really
close to
Canada.
Will you please not tell Homeland Security
about me?
DR. BORER: Thank you very much, Norman,
and thank you for staying way on time
because we
are now 17 minutes ahead of schedule,
which is
good.
The sponsor has a 90-minute presentation.
We'll try to allow you to move along as
well as we
can, but undoubtedly there will be some
clarification
questions. We ought to try to hold the
15
questions that we ask to clarification
issues
during the presentation, if we can, and
we can get
into the meat of the substantive
discussion
afterwards.
The presentation will be
introduced by Dr.
Cameron, the Vice President of Exanta.
DR. CAMERON: Thank you, Mr. Chairman,
members of the committee, ladies and
gentlemen,
good morning. I'm Dr. Hamish Cameron, the Vice
President of Exanta at AstraZeneca, and
with my
colleagues we're pleased to present
ximelagatran, a
new oral anticoagulant.
After a 20-year journey to
discover and
develop this new medicine and half a
century
without significant innovation in this
area of
therapeutics, we believe ximelagatran,
the first
oral treatment in the new drug class
direct
thrombin inhibitors is a real advance in
oral
anticoagulation.
Ximelagatran has a mechanism of
action
that's quite different from the vitamin K
antagonists like warfarin and can provide
the first
16
oral alternative to warfarin, today's
only option
for long-term anticoagulation.
Anticoagulation is the major
approach to
both the prevention and treatment of
thromboembolic
disease, a disease that's the final
common pathway
for many life-threatening conditions,
like stroke,
myocardial infarction, and pulmonary
embolism. And
it's the commonest cause of death and
disability in
America today.
At the outset we must ask: Given the
widespread availability of the vitamin K
antagonists like warfarin, why is there a
need for
a new oral anticoagulant? Warfarin is a highly
efficacious anticoagulant and one of the
top ten
most prescribed drugs, used in nearly
every medical
specialty by 3 million patients in the
U.S.
involving 32 million prescriptions every
year. But
it's been in the top five, sometimes
number one, in
the lists of drugs associated with
significant
interactions, medication errors, serious
bleeding,
and hospital admissions.
Warfarin's profile of
unpredictable
17
kinetics and dynamics; food, alcohol, and
multiple
drug interactions; together with its
acknowledged
narrow therapeutic index--too little
warfarin, and
there's the risk of residual clotting;
too much,
and the risk of bleeding--all these drive
the need
for a lifetime of INR coagulation
monitoring and
never-ending individual dose titration.
To put it simply, you don't get
the
benefit of warfarin from just taking the
tablet.
Its overall effectiveness is highly
dependent on
how it's managed. And it's this fact that frames
the innovation of ximelagatran.
Many patients and doctors fear
the risk of
bleeding that comes with unpredictable
anticoagulation.
This fear tends to result in
under-treatment, quite paradoxical in
high-risk
elderly patients, or in about half the
overall
patients eligible for warfarin, little or
no
treatment at all.
We started a discovery program
targeting
thrombin in 1985. We sought to develop a new oral
anticoagulant, an alternative to
warfarin, with a
18
profile that would allow fixed dosing
without
coagulation monitoring, further supported
by a low
potential for food and drug interactions.
We looked for a rapid onset and
offset of
action to simplify turning
anticoagulation on and
off, which is one of the challenging
aspects of
warfarin treatment. And all this had to be
achieved with an acceptable bleeding
profile.
Today, we believe these
objectives have
been met by the Ximelagatran Development
Program,
involving 82 clinical studies and
enrolling over
30,000 subjects. More than 17,000 people received
ximelagatran with 3,500 patients dosed
for over a
year.
And our longest patient exposure has now
reached five years.
Here are the three proposed
indications in
the
current NDA, spanning exposures from days to
several years. The first is the long-term
secondary prevention of venous
thromboembolism,
VTE, after standard treatment for an
acute episode.
Treatment of acute VTE involved six
months of
anticoagulation with warfarin, but at the
start of
19
this development in 1999, it was unknown
whether
longer treatment would be
beneficial. And so a
placebo-controlled study, THRIVE III, was
conducted. As you'll see, this study demonstrated
a highly significant reduction of VTE
during
longer-term prophylactic treatment. And as a
placebo-controlled study, it provides the
strongest
evidence of ximelagatran's antithrombotic
efficacy.
The second indication is the
prevention of
VTE after knee replacement surgery. Patients
without anticoagulant prophylaxis run a
high risk
of DVT and pulmonary embolism, and in the
U.S.,
warfarin is the most widely used drug,
started late
on the day of surgery to reduce this
risk. In two
warfarin controlled studies, EXULT A and
EXULT B,
we've shown a significant reduction of
VTE risk for
ximelagatran compared with warfarin.
The third indication is the
prevention of
stroke and other thromboembolic
complications
associated with atrial fibrillation. Here with the
large SPORTIF III and V trials, we've
shown the
efficacy of ximelagatran to be comparable
to
20
well-controlled warfarin.
Across these pivotal, mainly
outcomes-based studies involving
independent
endpoint adjudication, we've demonstrated
ximelagatran's antithrombotic efficacy
and recorded
a favorable leading profile, equivalent
to and in
some cases better than the
comparator. We detected
a
signal of raised hepatic enzymes with chronic
treatment, and so we've conducted a very
detailed
analysis, consulted with experts, and
believe the
risk can be adequately managed.
I should highlight that in your
briefing
packs and our safety presentation, we've
included
data from two other large studies in
other
indications, not for consideration today,
but which
contribute nearly 4,000 patients. The THRIVE
treatment study is the first pivotal
study looking
at initial VTE treatment, and the second
study is
soon to start, while the ESTEEM trial is
a Phase II
dose guiding study in the post-acute
coronary
syndrome setting.
These data enrich the overall
safety
21
assessment, including patients from very
different
clinical settings and a wide range of
characteristics, and
we're going to review all the key
data in the presentations that follow.
We believe ximelagatran with
its
predictable anticoagulant effects and
favorable
bleeding profile has a positive
benefit/risk in the
proposed indications, provided it's used properly.
And part of that proper use is the
introduction of
an appropriate risk management program
directed
towards the hepatic risk.
We made an initial proposal
with our
submission which had been developed with
extensive
external consultation and field testing,
but we
fully recognize, following the
deliberations of
this committee and further discussions
with FDA,
the program will need to be developed and
strengthened further before an approach
can be
finalized in the best interests of
patients. We
are committed to working with FDA to
achieve the
most appropriate risk management program
to ensure
the safe use of ximelagatran because
patient safety
22
is, has been, and always will be
AstraZeneca's top
priority when we introduce new medicines
into
clinical practice.
Since 1998, we've met
repeatedly with FDA
throughout ximelagatran's development--in
end of
Phase II meetings, a pre-NDA interaction,
and
there's a meeting coming up to discuss
the nature
and extent of the risk management
program. The NDA
was submitted in December 2003.
I should add that all the same
data are
now being reviewed in Europe by the
French agency
before a mutual recognition
procedure. But there's
one difference worth noting. Given the quite
different clinical practice regarding
anticoagulant
prophylaxis in orthopedic surgery,
separate
developments were conducted in the U.S.
and Europe.
The European program, reflecting local
practice and
starting treatment much closer to the
time of
operation, was the subject of a separate
earlier
regulatory submission and completed the
mutual
recognition procedure in May this
year. And the
first orthopedic launch was in Germany in
June.
23
Now, you have the data on this
program in
your briefing packs, but with the
significant
timing, comparator, and formulation
differences,
our presentations today will largely
focus on the
data directly relevant to the NDA
orthopedic
surgery application.
Here's the agenda for our
session. Dr.
Troy Sarich will review clinical
pharmacology; Dr.
Jay Horrow, efficacy; and Dr. Sunita Sheth,
safety;
allowing an overall evaluation of
ximelagatran in
the three requested indications. Dr. Jonathan
Halperin from Mount Sinai Medical Center
will then
give his views on the benefit/risk of
ximelagatran
in clinical practice. And throughout our
presentations, we hope to cover for you
all the
specific comments raised by the agency in
their
briefing document.
In addition to Dr. Halperin,
we're also
joined by other consultants: Dr. Gerald Faich, Dr.
Lloyd Fisher, Dr. Peter Kowey, and Dr.
James Lewis.
In summary, then, ximelagatran
is a new
oral anticoagulant that provides the
first
24
alternative to warfarin after 50
years. We believe
a total review of the available clinical
data
supports a positive benefit/risk in each
of the
proposed indications. Ximelagatran can enhance
health care delivery in America and
throughout the
world to help prevent a range of
debilitating and
life-threatening thromboembolic diseases.
Thank you. Now I'd like to introduce Dr.
Troy Sarich for clinical pharmacology.
DR. SARICH: Good morning.
I'm Dr. Troy
Sarich, Director of Clinical Pharmacology
at
AstraZeneca. I'll now present an overview of the
clinical pharmacology of ximelagatran in
which we
have performed both the traditional clinical
pharmacology studies and population
pharmacokinetic
analyses.
Ximelagatran is an oral direct
thrombin
inhibitor. It's rapidly bioconverted to the active
form, melagatran. The bioconversion, which is
Cytochrome P-450 independent, involves both
de-esterification and a reduction that
occurs
throughout the body.
25
The exposure to melagatran is
linear
across a dose range from 5 to 98 milligrams
ximelagatran. The pharmacokinetics are predictable
over time with repeated dosing, and the
elimination
half-life of melagatran is approximately
4 to 5
hours in patients. Once formed, melagatran is
primarily eliminated from plasma by a
glomerular
filtration.
Thrombin is a key enzyme in the
coagulation cascade. It converts fibrinogen to
fibrin, activates platelets, and induces
its own
generation. Melagatran directly inhibits thrombin
as a classic competitive and
reversible-binding
enzyme inhibitor. There's a direct relationship
between the pharmacokinetics and
pharmacodynamics
of ximelagatran. Its active when present in
plasma, and once eliminated from plasma,
its effect
is gone.
Preclinical investigations
indicated an
antithrombotic effect of melagatran at
approximately 0.05
micromolar, with increasing effect up
to approximately 0.5 micromolar.
26
In humans, ximelagatran
prolongs clotting
time assays. The thrombin time assay shown here
was prolonged in a linear manner at
concentrations
as low as 0.05 micromolar. In addition, melagatran
prolongs in a concentration-dependent
manner the
activated partial thromboplastin time,
although it
is less sensitive.
Additional investigations using
pharmacodynamic models in humans
demonstrated
evidence for inhibition of thrombin
generation
indicated by concentration-dependent
reduction and
thrombin-antithrombin complex levels and
platelet
activation indicated by
concentration-dependent
reduction in beta thromboglobulin levels
at
melagatran concentrations at or near 0.05
micromolar. All together, it was clear from these
data that direct inhibition of thrombin
by
melagatran resulted in the intended
anticoagulant
activity in humans.
After oral administration, the
inactive
pro drug ximelagatran is rapidly
eliminated from
plasma as it is biotransformed to
melagatran, shown
27
in blue, with peak melagatran
concentrations
occurring approximately 2 to 3 hours
post-dosing.
Melagatran plasma concentrations greater
than 0.05
micromolar are achieved early after oral
ximelagatran,
indicating a rapid onset of action which
simplifies the initiation of oral
anticoagulation.
Concentrations remain above 0.05
micromolar
throughout the dosing interval,
supporting a
twice-daily dosing regimen.
And as shown here, the rapid
onset of
action of oral ximelagatran is not
altered when
co-administered with food. Although there's an
approximately one-hour delay in the time
to C-max,
there is no effect on the AUC or Cmax of
melagatran.
Warfarin's well-recognized drug
interaction profile is largely related to
its
metabolism by the Cytochrome P-450 system
and its
high plasma protein binding. Ximelagatran is not
metabolized by and does not inhibit the
major
Cytochrome P-450 enzymes listed
here. It also has
low plasma protein binding, and along
with the
28
majority of melagatran eliminated from
plasma by
glomerular filtration, this leads to an
inherently
low potential for drug interactions.
Our investigations have
identified
pharmacokinetic interactions with erythromycin
and
azithromycin. Erythromycin results in an
80-percent or less than twofold increase
in
melagatran plasma levels, with a smaller,
40- to
60-percent increase with azithromycin.
These changes are within the
overall range
of melagatran exposures in patients, and
as
outlined in detail in your briefing
document,
investigation into the potential impact
of this
pharmacokinetic interaction found no
signal for
increases bleeding events or increased
ALT
elevations in the approximately 230
patients
receiving ximelagatran and macrolide
antibiotics in
the long-term studies. These data do not suggest
an important clinical impact of these
pharmacokinetic interactions.
We have conducted many other
interaction
studies where we've found no significant
29
pharmacokinetic interactions. As shown by the mean
melagatran AUC ration and the 90-percent
confidence
interval within or slightly outside the
0.8 to 1.25
no interaction interval. The drugs investigated
include alcohol, common cardiovascular
medications,
an NSAID, a sedative, and several
antibiotics.
These results are consistent with
population
pharmacokinetic analyses indicating a
lack of
interaction with commonly used
comedications in the
patient studies. Taken together, these data
suggest that ximelagatran has a low
potential for
drug interactions.
Melagatran is primarily renally
eliminated
from plasma, and so we've carefully
investigated
the impact of renal function on the
pharmacokinetics of
ximelagatran. In the three patient
populations under consideration today,
melagatran
exposure increases as calculated
creatinine
clearance decreases. For this reason, severe renal
impairment, a calculated creatinine clearance
less
than 30 mLs per minute, was an exclusion
criteria
for our clinical studies, and we're
currently
30
investigating an alternative dosing
strategy in
that population.
It's notable that we've
gathered
considerable experience with ximelagatran
in
patients with mild to moderate renal
impairment as
approximately 45 percent of the Phase III
patient
population had a calculated creatinine
clearance
between 30 and 80 mLs per minute. The median
exposures in these patients are about 1.5
to 2.5
times higher, respectively, than patients
with
normal renal function, but there's
considerable
overlap in melagatran exposure between
groups,
suggesting dose adjustment was not
necessary.
We've also studied the
potential effects
on the pharmacokinetics of ximelagatran
within
other special populations, and other than
differences in renal function between
groups, we
have not identified other important
effects of age,
gender, race, obesity--as measured using
body mass
index--or body weight on the
pharmacokinetics of
ximelagatran.
The agency has suggested there
should be a
31
dose adjustment for ximelagatran in
patients with
renal impairment given the higher levels
of
melagatran in these patients. But I'd like to show
you why a fixed dose, as used in our
clinical
studies, is appropriate across the
patient
populations studied.
We do agree with the agency's
assessment
that there's no need for dose adjustment
in
orthopedic surgery patients and that
there was no
increased bleeding related to melagatran
exposure
in VTE secondary prevention
patients. We do
acknowledge an association between
increasing
melagatran exposure and increasing
incidence of
major bleeding in atrial fibrillation
patients.
But this relationship appears confounded
by the
correlation between melagatran exposure
and the
age-related decrease in calculated
creatinine
clearance.
Shown here from the SPORTIF
trials is the
relationship between calculated
creatinine
clearance and major bleeding. As you can see,
major bleeding increased with declining
renal
32
function whether patients received
ximelagatran or
INR-controlled warfarin. This suggests the
increase in melagatran concentrations in
patients
with renal impairment is not associated
with
increased bleeding versus INR-controlled
warfarin.
It's also important to note
that stroke
risk increased with decreasing calculated
creatinine clearance, and the vast
majority of
these strokes were ischemic. So there's a
possibility that a dose reduction in
renally
impaired patients intended to decrease
bleeding may
increase the risk of stroke in those
patients at
highest risk.
We should also consider the
hepatic
findings, as will be presented by Dr.
Sheth, and we
have examined the possible relationship
between
melagatran exposure and ALT
elevations. As pointed
out in our briefing document, we have
observed an
association between increasing melagatran
exposure
and increasing ALT elevations greater
than 3 times
upper limit of normal, but this
relationship is
very weak. And as shown here, the relationship
33
between melagatran AUC and peak ALT
elevation in
individual patients, while statistically
significant, does not suggest a clear
relationship
between melagatran exposure and ALT
elevations.
In addition, we agree with the agency's
conclusion that, aside from the ALT
elevations
noted with ximelagatran, there is no
difference in
the overall adverse event profile between
ximelagatran and comparators in the
long-term
dosing study pool.
Factoring in the occurrence of
major
bleeding, stroke and systemic embolic
events, ALT
elevations, and the overall adverse event
profile,
the observation of increased plasma
melagatran
concentrations in renal impairment does not
appear
to justify a dose reduction in these
patients. We
believe our data support a fixed dose of
ximelagatran in the patient populations
studied.
Now I'd like to show the steady-state
plasma concentrations of melagatran in
atrial
fibrillation patients receiving a fixed
dose of 36
milligrams ximelagatran twice daily. There are
34
four key points here.
Plasma concentrations fluctuate
during the
dosing interval, remaining largely above
0.05
micromolar and infrequently exceeding 1
micromolar.
Mean trough melagatran
concentrations
after 36 milligrams are approximately 0.2
micromolar. So should a patient miss a dose of
ximelagatran, the 4- to 5-hour half-life
of
melagatran means that low but
pharmacologically
active concentrations remain for up to 24
hours
post-dosing. And the effect of melagatran is gone
once it is cleared from plasma by the
kidneys.
This emphasizes the importance
of
maintaining good diuresis in the
management of
bleeding.
And while there is no specific antidote,
if needed, melagatran can be dialyzed.
And, lastly, the APTT is
prolonged in
patients and may help identify a residual
anticoagulant effect.
A critical aspect of oral
anticoagulation
is maintenance of a stable effect over
time, and we
have confirmed the long-term stability of
oral
35
ximelagatran. Shown in yellow are the plasma
concentrations of melagatran in atrial
fibrillation
patients in the Phase II study, SPORTIF
II. We
remeasured plasma melagatran
concentrations in a
subset of those same patients between 13
to 16
months later in SPORTIF IV, a long-term
continuation study of SPORTIF II.
The mean plasma concentrations
of
melagatran are completely overlapping,
and the
variability in exposure within individual
patients
was low, with a coefficient of variation
of 25
percent, indicating that oral
ximelagatran results
in stable and reproducible plasma
concentrations of
melagatran with long-term repeated
dosing. This
stability enabled us to conduct our
clinical
studies using a fixed dose without
coagulation
monitoring.
So we can conclude from this
extensive
clinical pharmacology program
pharmacologically
active concentrations of melagatran are
rapidly
achieved and maintained in a broad range
of
patients.
There is also no effect of food or
36
alcohol and a low potential for drug
interactions.
The key attributes of
ximelagatran are,
therefore, its oral availability, rapid
onset of
action, low potential for drug
interactions, and
use at a fixed dose without coagulation
monitoring.
Now I'd like to introduce Dr.
Jay Horrow,
who will provide to you an evaluation of
the
efficacy of ximelagatran demonstrated in
Phase III
clinical studies.
DR. BORER: We'll just stop for one moment
to make sure there are no issues that
need to be
clarified. The relation of renal function to
melagatran exposure undoubtedly is going
to be
discussed to a greater extent later, but
I think we
should hold that until we hear from the
FDA
presentations, and then we can talk about
that.
But if there are any issues that need to
be
clarified regarding the pharmacology, we
should do
that now.
Steve?
DR. NISSEN: Two very brief questions. Is
anything known about the mechanism of
macrolide
37
interaction? Have you explored that at all?
DR. SARICH: Yes.
We were slightly
surprised to find that interaction since
we don't
interact with the P-450 system. It appears that
the interaction involves transport
proteins of some
kind, and we've looked at a range of
different
compounds that we've investigated, and it
at this
point appears isolated to the macrolide
antibiotics
we've studied.
DR. NISSEN: And the second question is:
You showed the coagulation effect during
therapy,
and I wondered if you have additional
data on what
happens in, let's say, the first 72 to 96
hours
after terminating therapy. Is there evidence of a
rebound phenomenon?
DR. SARICH: We have not observed that
pharmacologically, as far as coagulation
time
assays.
DR. NISSEN: Okay.
But that has been
studied.
DR. SARICH: We've followed out to 24
hours after single-dose administration
and not seen
38
any evidence--
DR. NISSEN: But not longer than 24 hours?
DR. SARICH: Not that I can recall.
DR. BORER: John?
DR. TEERLINK: The other question I have
is:
In terms of the relationship between the
melagatran AUC versus the peak ALT elevations,
how
was the melagatran AUC derived?
DR. SARICH: Yes, these were derived using
a population pharmacokinetic model. So the
patients that received ximelagatran in
the Phase
III clinical studies had plasma samples
collected.
Over 80 percent of the Phase III patient
population--in the long-term population
had a
plasma sample collected. Using a pharmacokinetic
model that was developed by the team, we
were able
to estimate the exposure to melagatran in
those
patients.
DR. BORER: Ron Portman?
DR. PORTMAN: Noting differences in the
chronopharmacology of drugs, were the
curves you
showed similar for both the morning and evening
39
doses?
DR. SARICH: Are you speaking about the
coagulation time assay--
DR. PORTMAN: No. I
was talking about the
plasma concentrations.
DR. SARICH: Pharmacokinetics?
DR. PORTMAN: Right, pharmacokinetics.
DR. SARICH: Yes, they are consistent
under administration during the day or
overnight.
DR. BORER: Jonathan, go ahead.
DR. SACKNER-BERNSTEIN: In the analysis
that you showed the stability of the
concentrations
of the drug over time from the SPORTIF II
and
SPORTIF IV population, did you perform
that
analysis restricting to patients who had
samples at
both times? Because the analysis you showed had a
larger population at baseline compared to
a subset
later.
DR. SARICH: Right.
We've done it both
ways.
The figure actually represents the larger
number in the SPORTIF II study and a
smaller number
in SPORTIF IV. The intra-subject variability I
40
noted was only the subjects that had
sampling at
both time occasions.
DR. BORER: Alan?
DR. HIRSCH: In the PK and AUC curves that
you've generated, were there any changes
or
differences noted based on ethnicity,
geographic
sampling of a population, or gender?
DR. SARICH: Are you asking pharmacokinetic--
DR. HIRSCH: Yes, PPK differences between
subgroups.
DR. SARICH: The main factor we've
observed between any subgroups has been
differences
in renal function, calculated creatinine
clearance.
We have not observed any significant
effects of
other demographic parameters, age,
gender, race,
BMI, body weight. It appears that exposure--the
most influential demographic factor is
calculated
creatinine clearance.
DR. BORER: Susanna?
DR. CUNNINGHAM: Did you have a sufficient
African American population to actually
know
41
anything about what the African American
area of
the curve might be or handling of the
drug?
DR. SARICH: We have performed pharmacokinetic
studies in that population. I should say
both--I'll show you some data here from a
small
study.
It's not African Americans per se, but it
was a study in Europe, in Paris, in fact,
where we
had 12 blacks, 12 Asians, and 12
Caucasians, and
found no real differences between these
groups.
If we looked at the entire
patient
population, we can see here--if we look
at--you can
see the Caucasian population here. There's over
6,000 patients. The blacks where we had
appropriate pharmacokinetic information,
were 115,
as well as Asians, and the category of
other, and
no differences between these populations.
DR. BORER: Tom?
DR. PICKERING: Do you have any data on
interaction with aspirin?
DR. SARICH: Yes, we have performed
actually two studies with aspirin. There's no
pharmacokinetic interaction with
aspirin. We see
42
an additive effect on the capillary
bleeding time,
which is somewhat expected.
DR. BORER: Beverly?
DR. LORELL: Yes, with regard to body
size, you commented on and emphasized
obesity.
What about the other end of the scale,
very small
body size? Sometimes an issue in elderly women who
might be candidates for several of these
indications.
DR. SARICH: We have less data in very
small individuals, but what we know about
that
population is that it's primarily their
calculated
creatinine clearance that influences
their
exposure.
DR. BORER: Jonathan?
DR. SACKNER-BERNSTEIN: I know we're going
to get back to the renal function
question, but
there was one set of slides you showed
where you
tried to give us some reassurance about
the
relationship between bleeding and renal
function.
And you showed the risk of bleeding as
calculated
creatinine clearance reached the low end
of the
43
spectrum.
I wonder if you performed any
sort of
retrospective power calculation on your
ability to
detect a difference in risk, in
particular in the
patients who we may be likely to see
treated with
this drug in clinical practice, those
over 70, over
75, where calculated creatinine
clearances often
are in the 40s. So do you have an analysis there
between 30 and 50 with conditional power
to
actually detect a difference in bleeding
risk
there.
DR. SARICH: I think we could probably
best address that after the
presentations. We do
have data there, and rather than getting
into that
discussion, if the Chair would agree, we
could
address that, bring an answer to you for
that.
DR. BORER: Is that okay, Jonathan?
DR. SACKNER-BERNSTEIN: Yes.
DR. BORER: Okay.
Thank you.
T1B DR. HORROW: Ladies and
gentlemen, I'm Dr. Jay Horrow from
AstraZeneca. We
will now present Phase III data
demonstrating that
44
ximelagatran is an effective oral
anticoagulant.
In the first indication,
long-term
secondary prevention of venous
thromboembolism, we
will show ximelagatran superior to
placebo. In the
second indication, prevention of VTE
after total
knee replacement, ximelagatran was
superior to
well-controlled anticoagulation with
warfarin. And
in the chronic prevention of stroke,
ximelagatran
was noninferior to warfarin.
These indications represent a
broad range
of patient populations. We'll begin with the first
one:
secondary prevention of VTE.
Evidence has been accumulating
that
patients with acute VTE benefit from prolonged
anticoagulation after acute
treatment. The THRIVE
III trial comparing ximelagatran to
placebo
contributes to this growing body of
evidence.
Randomized patients had an acute symptomatic
VTE
objectively confirmed and had completed 6
months of
treatment without VTE recurrence, also
objectively
documented at randomization. Anticoagulation was
desirable but not essential for these
patients,
45
that is, they had idiopathic VTE or
probable
hypercoagulable conditions. Health status had to
be compatible with survival for an
additional 18
months.
In THRIVE III, 1,223 patients
receives in
double-blind fashion either oral
ximelagatran 24
milligrams twice daily or placebo for up
to 18
months.
Selection of 24 milligrams for this trial
came from a consideration of preclinical data
and
data from Phase II trials in the
orthopedic surgery
indication. These PK data from a Phase II European
trial in patients undergoing hip or knee
replacement demonstrate that
administration of 8
milligrams ximelagatran twice daily, the
lowest
curve, achieves plasma melagatran
concentrations of
about 0.05 micromolar. This is the level at which
anticoagulant activity with melagatran
begins based
on the data previously shown by Dr.
Sarich.
Progressively higher doses of
oral
ximelagatran, 12, 18, and 24 milligrams,
achieved
higher melagatran concentrations, more
anticoagulant
activity, and more time above the 0.05
46
micromolar threshold for each dose.
Outcome data from that same
orthopedic
surgery trial suggest that 24 milligrams
is the
most promising dose for efficacy. The 24-milligram
dose also had a reassuring bleeding
profile. We
chose 24 milligrams for THRIVE III with
placebo
comparator without establishing
dose-limiting
toxicity, in this case bleeding. The choice was an
informed judgment taking into consideration,
first,
the need for efficacy demonstrated by the
benefits
seen here in joint replacement, an
intense
thrombotic stimulus; and, second, the
need to avoid
excess bleeding because the standard of
care is no
anticoagulant therapy at all.
The trial compared ximelagatran
to placebo
in the rate of recurrence of symptomatic,
objectively confirmed VTE. VTE encompasses both
deep vein thrombosis, DVT, and pulmonary
embolism,
PE, because PE originates from a thrombus
in the
systemic venous circulation, whether
overt or not.
The primary endpoint compared
ximelagatran
to placebo using a time-to-event
analysis. A
47
recurrence of VTE required signs or
symptoms of
VTE, that is, a clinical event, and
subsequent
objective confirmation. A blinded independent
endpoint committee evaluated and
adjudicated all
clinical endpoints, including major
bleeding
events.
The ximelagatran- and
placebo-treated
cohorts displayed similar demographic
profiles. As
indicated by creatinine clearance between
30 and
80, 23 percent had some degree of renal
impairment.
The index VTE event was or included
pulmonary
embolism for more than one-third of
patients. This
Kaplan-Meier curve shows the cumulative
incidence
of the primary outcome in the
ximelagatran and
placebo groups, analyzed by intention to
treat.
Seventy-one patients in the placebo group
suffered
recurrent VTE, including 23 PEs, for a
cumulative
rate of 12.6 percent, while only 12
patients in the
ximelagatran group had recurrent VTE,
including
only two PEs, for a cumulative rate of
2.8 percent.
The 9.8-percent difference, significant
at p less
than 0.0001 by log rang test indicates
that one VTE
48
recurrence is prevented by ximelagatran
treatment
for up to 18 months to 10 patients. The associated
hazard ratio, 0.16, indicates a risk
reduction of
84 percent by ximelagatran relative to
placebo.
The composite endpoint of total VTE
included both DVT and PE. Benefit of ximelagatran
over placebo occurred for each component
of this
composite endpoint--clinical DVT,
clinical PE, and
their combination.
The superiority of ximelagatran
to placebo
is robust to multiple, prespecified
sensitivity
analyses listed here. Each comparison demonstrated
a significance level less than 0.0001
Here we examine efficacy in
subpopulations. Small
diamonds depict point estimates of
the odds ratios of ximelagatran to
placebo, and
horizontal bars show their 95-percent
confidence
intervals. Superiority of ximelagatran over
placebo remains in all subgroups strata
of
reasonable size.
In THRIVE III, the oral
thrombin inhibitor
ximelagatran, 24 milligrams twice daily
for up to
49
18 months, effectively reduced the number
of
recurrent VTE events following 6 months'
treatment
of an acute VTE. The results are robust and
consistent across multiple endpoints and
subgroups
and demonstrate a clinically relevant
benefit.
The second indication under review today
is the prevention of VTE in patients
undergoing
knee replacement surgery. Major joint replacement
surgery challenges any anticoagulant to
prevent VTE
without counteracting surgical hemostasis. VTE
prevention contributes heavily to the
benefit/risk
balance for joint replacement surgery.
The current options to reduce
the
occurrence of VTE after total knee
replacement
include the injectable agents
low-molecular-weight
heparin and fondaparinux and oral
warfarin. One
FDA comment regards the choice of
warfarin as the
comparator for these trials.
We chose warfarin for several
reasons:
First, it is the agent most
commonly used
for this purpose in North America, and we
administered warfarin, as orthopedic
surgeons do,
50
beginning the night of surgery.
Second, warfarin, like
low-molecular-weight heparin, is a Grade
1A
recommended therapy for this purpose,
according to
current American College of Chest
Physicians
Consensus Conference Guidelines.
And, third, warfarin is
associated with
less bleeding than the injectable
anticoagulants
and so is a more daunting comparator for
ximelagatran in terms of surgical
hemostasis.
Two independent double-blind
Phase III
trials--EXULT A and EXULT B--enrolled
patients
undergoing primary elective total knee
replacement.
EXULT A studied 24 and 36 milligrams
ximelagatran
and warfarin. We had studied 24 milligrams in this
context previously and found protection
similar to
but not better than warfarin at p equal
0.07. We
found that result surprising. Unsure whether or
not it was a Type II error, we designed
EXULT A
with two ximelagatran arms: one using 24
milligrams and the other using 36
milligrams.
Warfarin and its paired placebo
began, as
51
typically practiced in the U.S., the
evening of the
day of surgery while ximelagatran and its
placebo
began early on the morning after the day
of
surgery.
Treatment continued for 7 to 12 days,
after which all patients underwent
bilateral
venography.
Based on the results of EXULT
A, EXULT B
studied only 36 milligrams ximelagatran
and
warfarin.
Warfarin was aggressively and
successfully dosed to drive the INR
rapidly to its
target of 2.5, with an accepted range of
1.8 to
3.0.
The primary outcome formed the composite of
distal and proximal DVT by venogram
performed
between days 7 and 12, objectively
confirmed
symptomatic DVT or pulmonary embolism up
to 2 days
after venography, and all-cause mortality
up to 2
days after venography. Both trials utilized the
same blinded independent committee for
event
adjudication.
The treatment groups in the
EXULT trials
were balanced and represented well the
population
of patients in the United States
undergoing total
52
knee replacement. More than a third of the cohort
displayed some degree of renal
impairment.
Here are the primary results
for EXULT A
and EXULT B. In EXULT A, ximelagatran 36
milligrams, in yellow, showed superiority
to
well-controlled anticoagulation with
warfarin, in
gray, at p equals 0.003. EXULT B confirmed those
results, with p less than 10
-5. These results yield
relative risk reductions of 26 and 29
percent and
numbers needed to treat of 14 and 11,
respectively.
In EXULT A, ximelagatran 24
milligrams, in
orange, and warfarin, in gray, did not
differ, with
event rates of 24.9 and 27.6 percent,
respectively.
That p value is 0.28.
The delay in anticoagulation
with warfarin
administration suggests that it may act
like a
placebo in EXULT. In fact, warfarin rates, in
gray, are the lowest ever obtained in
knee
replacement clinical trials with
warfarin, perhaps
because of the rapid achievement in EXULT
of
therapeutic INRs. Placebo rates are historically
over 60 percent, and the mean INR in
EXULT was 2.4
53
on post-op day 3. The warfarin group provided a
formidable comparator for ximelagatran in
the EXULT
studies.
Here we see results for the
components of
the composite primary endpoint. As expected, the
majority of events occurred in the distal
leg.
Rates for proximal DVT, for PE, and for
death were
low in all treatment groups. Another point raised
by FDA is how clinically relevant distal
DVT is as
a component of that endpoint. It's important to
note that 10 to 20 percent of distal
thrombi extend
to become proximal thrombi, and either
one can
cause pulmonary embolism, making all
three
phenomena clinically relevant components
of a
composite endpoint. In fact, proximal and distal
deep vein thrombosis detected by
venography,
whether symptomatic or not, is a primary
endpoint
historically accepted by the agency for
VTE
prophylaxis registration trials.
This display of the primary
outcome by
subgroup strata shows differences in
event
incidences between the pooled
36-milligram
54
ximelagatran and pooled warfarin
groups. These
results, with small numbers in just a few
subgroups, reveal no discrepancies in
efficacy in
any particular subpopulation.
Oral ximelagatran, 36
milligrams, provided
superior protection against VTE and
all-cause
mortality compared with well-controlled
anticoagulation with adjusted-dose
warfarin, a
clinically relevant comparator. This superior
protection was consistent across multiple
subgroups. These data support the efficacy of
ximelagatran for the indication
requested.
The third indication considered
today is
the protection of patients from stroke
and other
thromboembolic complications of atrial
fibrillation. We have demonstrated that
ximelagatran provides this protection, as
well as
does warfarin, across a broadly based
patient
population.
Two independent pivotal
trials--SPORTIF
III, dosed, open-label in 23 countries in
Europe
and Asia, and SPORTIF V, conducted
double-blind in
55
North America--enrolled patients eligible
for
warfarin therapy according to existing
treatment
guidelines, that is, those with
nonvalvular atrial
fibrillation with at least one additional
risk
factor for stroke. Each SPORTIF trial by itself
studied more patients than all previous
trials of
stroke prevention in atrial fibrillation
combined.
Each trial compared 36 milligrams twice
daily
ximelagatran to dose-adjusted warfarin in
preventing all strokes and systemic
embolism, hard
clinical endpoints in an
intention-to-treat
fashion.
The choice of 36 milligrams
came from
several considerations. There is no surrogate
marker for stroke and systemic embolism,
and both
events are devastating. Thus, we performed a
dose-ranging study for safety and
tolerability of
ximelagatran 20, 40, and 60 milligrams in
SPORTIF
II, a 3-month Phase II atrial
fibrillation study.
While the numbers were small in that
study,
bleeding was most frequent with 60
milligrams and
also the warfarin comparator, and less
frequent
56
with 20 or 40 milligrams
ximelagatran. We knew
that 24 milligrams was effective in the
Phase II
European orthopedic surgery program and
reasoned
that any downside impact of potential
additional
strokes with 24 milligrams would be far
worse than
the bleeding seen with 36 milligrams in
this
nonsurgical context. Using this educated judgment,
we chose 36 milligrams in the Phase III
atrial
fibrillation program.
Let's take a moment to consider
the
open-label nature of the SPORTIF III
trial. The
majority of prior stroke prevention
trials in
atrial fibrillation also utilized an
open-label
format based on the difficulty of
managing
anticoagulation in blinded fashion. SPORTIF III
featured open-label dosing at sites, but
also
centrally randomized allocation and two
additional
levels of blinding: blinded local assessment of
primary endpoints by study-affiliated
neurologists,
and blinded independent central committee
adjudication of all study endpoints. SPORTIF V
featured double-blind, double-dummy
medication, and
57
for patients receiving ximelagatran and
placebo
warfarin, sham INR values that mimicked
those
obtained during warfarin therapy.
The established efficacy of
warfarin
precluded a placebo comparison. Because warfarin
is so efficacious, it is reasonable to
establish
ximelagatran efficacy in comparison to
warfarin,
and we did so using a noninferiority
design. In
consultation with an executive steering
committee
and data safety monitoring board compose
of leaders
of prior stroke prevention trials and a
statistician expert in noninferiority
trials, we
prespecified a 2-percent per year
absolute
noninferiority margin. The choice of this margin
has been questioned. The choice of 2 percent arose
partly from an expected 3.1 percent
warfarin rate,
but more importantly, from consideration
of the
clinically tolerable absolute difference
in stroke
rates considering warfarin's overall
clinical
profile.
A similar consideration drove designers
of the SPAF III trial to power that trial
to detect
a 2-percent per year event rate with
upper
58
confidence bounds of 3 for a population
at lower
risk of stroke. Even so, we prespecified a more
conservative 2-percent upper confidence
limit. The
point estimate of the difference in event
rates
needs to be much smaller than 2 percent
for the
worst case, that is, the upper confidence
limit, to
be less than 2.
The strength of the 2-percent
per year
absolute margin resides in its clinical
relevance,
its prespecification, and that it is
conservative.
At screening, those patients already
taking oral anticoagulants interrupted
that therapy
to decrease INR to 2 or less by the time
of
randomization, at which time patients
received
either warfarin or ximelagatran. Each trial
achieved a degree of warfarin control
rarely found
in routine clinical practice. The warfarin-treated
groups constituted formidable comparators
for
ximelagatran, particularly in SPORTIF
V. Samsa and
colleagues found that most patients
taking warfarin
spend more than half the time on
treatment outside
the therapeutic range. In SPORTIF V, only 15
59
percent of patients did so.
The ximelagatran- and warfarin-treated
cohorts displayed nearly identical
demographic
profiles in each independent Phase III
trial, seen
here as pooled data. Patients reflected well the
elderly population of nonvalvular atrial
fibrillation patients requiring
anticoagulation for
stroke prophylaxis, and the majority had
impaired
renal function.
In SPORTIF III, warfarin, shown
in gray,
displayed an event rate of 2.3 percent
per year
compared to 1.6 percent per year with
ximelagatran,
shown in yellow.
In SPORTIF V, the rates were
1.2 for
warfarin, in gray, and 1.6 for
ximelagatran, in
yellow.
Primary event rates with
ximelagatran are
nearly identical in SPORTIF III and
SPORTIF V. For
warfarin, the rates fall within the range
of event
rates in previous trials, 0.6 to 4.1
percent per
year.
For comparison, the pooled rate
in prior
60
stroke trials for patients in this risk
category
taking placebo or aspirin was over 8
percent per
year.
The difference in event rates in SPORTIF
III, 0.66, favoring ximelagatran, had an
upper
confidence limit of 0.13, less than the
prespecified 2-percent margin. In SPORTIF V, the
difference of 0.5 favoring warfarin had
an upper
bound of 1.03, also less than the
prespecified
2-percent per year margin. Thus, each trial
independently succeeded by satisfying the
prespecified noninferiority criterion for
the
primary outcome.
As expected, most of the events
in this
composite outcome were ischemic strokes.
Hemorrhagic stroke and systemic embolism
occurred
more rarely and did not influence the
primary
endpoint substantially.
Several sensitivity analyses
confirmed the
results of the primary analysis. One such
analysis, depicted here, included
all-cause
mortality in the primary endpoint at the
suggestion
of the agency. SPORTIF III returned event rates of
61
4.2 and 5.1 for a difference of 0.87
favoring
ximelagatran, while SPORTIF V rates were
nearly
identical at 4.7 and 4.8.
Adding all-cause mortality
shifted each
study's event rate difference point
estimate to the
left in favor of ximelagatran.
Another sensitivity analysis,
depicted
here, used an on-treatment approach using
the same
endpoints and the same population, but
not counting
events that occurred after stopping study
treatment
for 30 continuous or 60 total days. The upper
bound of negative 0.18 indicates
superiority in
SPORTIF III. The value of 1.2 in SPORTIF V
indicates noninferiority to
well-controlled
warfarin.
For each trial, we also performed a
paper
comparison of ximelagatran to placebo by
factoring
in the results of the six prior stroke
prevention
trials.
We obtained original data from those
trials to utilize the same endpoint
events as in
SPORTIF.
Demographics of patients in these trials
were similar to those of SPORTIF
patients.
62
In these calculations, SPORTIF
III and
SPORTIF V separately demonstrated
statistically
significant risk reductions for
ximelagatran
relative to putative placebo, as did the
pooled
SPORTIF data. Ximelagatran works as an
anticoagulant in this population.
As before, here we see differences in
primary event rates according to
demographic
subgroups. These pooled results reveal no
discrepancies in any particular
subpopulation,
including the elderly, women, the obese,
and those
with poor renal function.
In conclusion, for atrial
fibrillation
each of two trials independently met its
objective,
demonstrating that 36 milligrams of
ximelagatran
taken twice daily prevented stroke and systemic
embolism to an extent similar to that of
well-controlled anticoagulation with
warfarin.
For long-term secondary VTE
prevention,
the THRIVE III trial demonstrated that 24
milligrams ximelagatran twice daily
prevented VTE
recurrence compared to placebo.
63
And in total knee replacement
surgery, the
two independent EXULT trials showed that
36
milligrams twice daily prevented VTE and
all-cause
mortality better than dose-adjusted
warfarin.
Based on five pivotal trial,
each the
largest in its field, involving more than
12,000
patients, these data establish the
effectiveness of
ximelagatran as an oral anticoagulant in
a variety
of patient populations at high risk for
thromboembolism.
Dr. Sunita Sheth will next
address
particular safety aspects of
administration of
ximelagatran for these indications.
DR. BORER: Thank you very much, Jay.
Again, we'll take a minute to
see if
anyone has any issues that require
clarification.
Clearly, we are going to talk about or
probably
we're going to be talking about the selection
of
the delta for the noninferiority trial,
but I don't
want to get into that discussion
now. We have some
extraordinary statistical fire power here
between
Tom on the committee and Lloyd Fisher and
Jerry
64
Faich sitting over there and the FDA
statisticians.
I think we'll wait on that until after
all the
presentations, including the FDA
presentations,
have been made. But if we have any issues of fact
that need to be clarified now, let's do
it.
Jonathan?
DR. SACKNER-BERNSTEIN: In the FDA
briefing document, it points out that
there were
patients who were withdrawn from the study for
whom
there is not information about whether
they
underwent or suffered any events. If that's
correct, please clarify, because it looks
as though
from the study flow that that means in
SPORTIF V as
many as 15 percent of the patients we
basically
would not have any clinical outcomes data
available
from the point in time when they
withdrew. Is that
correct?
DR. HORROW: We followed up on all of our
patients in the SPORTIF III and V trials
to the
greatest extent possible, and, in fact,
after we
were done with our follow-up, at the time
of final
closure, locking the database, we were
left, out of
65
7,922 patients, with only 63 patients
about whom we
were unsure of their final status.
DR. SACKNER-BERNSTEIN: So that would mean
that the FDA briefing document is incorrect,
because the FDA briefing document states
that--and
I'm looking at page 36 of the clinical
review from
Cardio/Renal Division. It says in the first
paragraph that patients that were
discontinued from
study medication and withdrew from study
were not
followed for primary efficacy endpoints
or death.
And then as you turn to page 45 with the
patient
disposition in SPORTIF V, it looks as
though
there's 300 study withdrawals from the
ximelagatran
group and 286 from the warfarin
group. So that
means that about 15 percent would have
incomplete
clinical outcomes data, but you're saying
there's
only 63.
So could you explain for us
where the
disparity should be settled?
DR. HORROW: It is conceivable that there
is a misinterpretation of the term
"study drug
discontinuation" and
"withdrawal from study." More
66
likely, the misunderstanding may accrue
from the
follow-up efforts that we made to
ascertain the
vital status of every patient in the
SPORTIF
trials.
We followed up on every
patient, aside
from the 63 that I just mentioned, and
are
confident in their vital status, knowing
whether
they were alive or dead, whether they had
a stroke
or not, in our database.
DR. FLEMING: Could I just clarify? So I
assume what you then did is you defined a
date of
data lock or closure where on that
calendar date
you wanted to follow all patients
relative to their
survival status and stroke status. Are you saying
then for all but 63 patients you knew
their
survival status and stroke status as of
that
calendar date for data lock?
DR. HORROW: Exactly, and that would be
the data lock date for each respective
trial--SPORTIF III and SPORTIF V. That is correct.
DR. BORER: Steve?
DR. NISSEN: I want to make sure I
67
understand how you maintained the blind,
particularly in SPORTIF III. Obviously with
warfarin, you may require frequent dose
adjustments
and so on. So in the open-label, particularly
SPORTIF III, how did you maintain--in
both trials,
I'd like to understand the procedures
that were
undertaken. I guess in the open-label trial there
was no blinding, right? The physicians and
patients knew what they were receiving;
is that
correct?
DR. HORROW: In the open-label trial?
DR. NISSEN: Yes.
DR. HORROW: It was open-label dosing, and
so you are correct that the physicians
and the
patients knew the drug, and the
evaluators, the
neurologists locally, and the central
adjudication
committee were blinded and didn't know.
DR. NISSEN: Okay, I understand. And
SPORTIF V, then, how did you adjust
warfarin and
maintain the blind? Explain to me how that was
done.
DR. HORROW: It was quite tricky and
68
involved quite a bit of work on the basis
of the
investigators and quite a burden for the
patients.
In SPORTIF V, all of the INR values were
obtained
in almost all cases by only two
laboratories--that's an incredible degree
of
standardization for
thromboplastin--either the
centralized laboratory or a point-of-care
machine
that had standardized cards.
In each case--well, for the point-of-care
machine, a coded number was produced by
the
machine.
That was called in to a central
randomization area, and that service then
faxed to
the site either the true INR value if the
patient
was really in the warfarin group or a
shammed INR
value if the patient was truly taking
ximelagatran.
So the site was unaware when it received
the fax
what group the patient was in.
If the test was done at the
centralized
laboratory, then the centralized
laboratory
likewise sent the results to the IBRS
site, the
specialized service, which then, again,
faxed
either the shammed or the true INR value
to the
69
site.
DR. NISSEN: And then dose adjustments,
how were those then made? I mean, obviously some
of the patients needed a dose adjustment,
so what
happened then?
DR. HORROW: Well, as you know, for
ximelagatran or its placebo there were no
dose
adjustments. But for warfarin or its placebo, each
investigator adjusted the dose based on
their usual
practice considering the patient and the
INR value
or shammed value--they didn't know which
it
was--they'd received by fax.
DR. NISSEN: So there was no--it was all
done per local physician practice. There was no
standard applied to how dose adjustments
were made.
Is that right?
DR. HORROW: That's correct. We did not
require all the investigators to adjust
their
patients' warfarin doses against some
standard.
This was to be a very real-world--as much
as we
could--type of adjustment in terms of
warfarin or
shammed dosing.
70
DR. NISSEN: And I assume the reason you
didn't do that in SPORTIF III was that
you just
felt it was too difficult.
DR. HORROW: In SPORTIF III, the
investigators were very uncomfortable
with blinded
anticoagulation testing and were
unwilling to move
forward in that regard.
In SPORTIF V, our North
American
investigators embraced the randomization
somewhat
more willingly.
DR. NISSEN: So you tried to do SPORTIF
III blinded but they wouldn't go along
with it? I
don't understand exactly what happened.
DR. HORROW: It was not possible
to get
the investigators in SPORTIF III to move
forward
with the blinded testing and
anticoagulation.
DR. NISSEN: You attempted it, and then
they weren't able to comply. Is that what
happened?
DR. HORROW: At an investigators meeting,
there was--
DR. NISSEN: A rebellion.
71
DR. HORROW: There was no support.
DR. BORER: Okay.
Bill? And then we have
Tom and Alan and John.
DR. HIATT: A comment and a question. In
the knee replacement studies, you
commented that
you achieved a rapid increase in INR and
that it
was 2.4 at day 3. And I just want to comment that,
you know, there's an association between
antithrombotic and anticoagulant effects
of
warfarin.
It takes 4 to 6 days for Factor II to be
depleted, so that's a false sense of
security
around the measurement of the INR. They're still
not antithrombotic.
So my question is: If you take the
three-quarters of patients at the end of
that study
who were, quote, therapeutic versus the
one-quarter
that were not, did you look at a subgroup
analysis
around difference in VTE rates at the end
of that
time?
Were the patients who were, in fact,
therapeutic by that number equivalent in
terms of
VTE rates compared with the patients who
were
sub-therapeutic?
72
DR. HORROW: My understanding of the
question is did we perform a subgroup
analysis near
the end of the treatment interval
regarding
patients--or based on the actual INRs of
the
patients.
We do not have that analysis.
DR. HIATT: I think the speculation would
be that the differences would be erased
in those
who were therapeutic, and a major
difference
between treatments would have been in
those who
were sub-therapeutic. That was my question.
DR. HORROW: This is quite possible, and
it's important to understand that the
EXULT trials
mimicked warfarin administration in the
orthopedic
surgery realm as it is currently
practiced today in
the United States. And so it was a very relevant
way to look at the effects.
DR. BORER: Tom?
DR. PICKERING: Can you tell us how the
INR control rates in the SPORTIF trials
compared
with the same rates in the six
warfarin-versus-placebo trials?
DR. HORROW: The INR rates in the six
73
index trials had somewhat of a spread, as
would be
expected, and it's actually possible to
see that as
the INR rates are better in some of those
trials,
so are the results in terms of the
decrease in the
warfarin event rate. And our results for INR
control were really quite in the middle,
2.5,
2.4--could we have the previous slide,
please? I
can show you some data on them.
This would be for SPORTIF V,
summary
statistics. Please note in the middle column
labeled ximelagatran, we are looking at
shammed
values, and you will note that we have
2.5 at 3
months for ximelagatran and 2.4 for
warfarin, at 12
months similarly, at 24 months
similarly--right in
the middle of the desired interval. And, of
course, the other thing that you might
note here is
that there is a threshold of 4.0 for the
shammed
values to ensure that no shammed value ended
up
putting a patient unnecessarily in the
hospital
because of an elevated shammed INR.
Nevertheless, as you can see by
the ranges
here, it's quite clear that the
investigators would
74
be unable to determine whether a patient
were in
one group or the other.
DR. PICKERING: That really wasn't my
question.
I was asking if there are comparable
data for the six warfarin placebo trials.
DR. HORROW: I don't have those data
available to show you at this time.
DR. BORER: I think they're in one of our
two books, Tom.
DR. HORROW: I believe they may be in the
briefing document.
DR. BORER: If I remember correctly, they
do show a fairly wide range, as you might
expect,
but we can get those data.
Okay. Alan?
DR. HIRSCH: I have two questions. One is
to follow up Steve's question regarding
the SPORTIF
III blinding. I just always believe it's terribly
important to have blinding as a component
of major
pivotal trials, acknowledging that lack
of blind
can really alter outcomes in unexpected
ways. So I
want to just run this through one more
time.
75
Pitying the investigators that
would not
go along with your request, the patients
knew their
study assignment, correct?
DR. HORROW: In SPORTIF III.
DR. HIRSCH: In SPORTIF III.
DR. HORROW: That's correct.
DR. HIRSCH: The physicians--
DR. HORROW: The patients knew their
assignment, as did the principal
investigators.
DR. HIRSCH: And coordinators.
DR. HORROW: That's correct.
DR. HIRSCH: So how would we have any
confidence that the adjudicating
neurologist would
have any blind maintained at all?
DR. HORROW: Well--
DR. HIRSCH: I worry.
DR. HORROW: Your point is well taken that
that cannot be assured with
certainty. We can say
that there were efforts made to make sure
that the
neurologist was not told on purpose the
assignment
of the patient, and we know also that all
members
of
the central adjudication committee, which
76
evaluated all the endpoint events upon
which the
results are based, did so in a totally
blinded
fashion.
DR. HIRSCH: I guess if there was
concordance between those two groups, I'm
somewhat
satisfied.
Let me come back with a
follow-up question
for EXULT, if I could. The data that we have
demonstrates benefits of ximelagatran
versus
Coumadin preventing DVT in this
population at risk
after total knee replacement. And as we'll discuss
later, most of that data is regarding
distal DVT,
which I do care about. But in the database, do we
have any evidence, quality-of-life
measurements,
girths, anything that demonstrates a
clinically
relevant effect for the patient? In other words,
in the absence of venographic
surveillance, would
the patient know there was a difference
in outcome?
DR. HORROW: I'd like to ask Dr. Scott
Berkowitz, who is the medical director
for that
particular trial, to address that
issue. Dr.
Berkowitz?
77
DR. BERKOWITZ: Hi.
Scott Berkowitz,
AstraZeneca. There was not any type of
quality-of-life assessment in this
short-term
trial.
The symptomatic events were collected as
well, including distal, proximal, and
PEs. They
were low, as they are in TKR trials and
did not see
a difference, a statistical difference.
DR. BORER: There was a question I was
going to hold until the end, but it seems
to be
relevant right here in view of Alan's point. You
probably have a back-up slide, and Alan
just
suggested that he probably has the data
off the top
of his head. But can you tell us, among people
historically from older trials where data
would be
available who have asymptomatic distal
DVT and who
aren't treated, what's the risk of
subsequent
thromboembolic events during some
follow-up period?
DR. BERKOWITZ: Well, we don't have the
greatest data on that, unfortunately, in the
literature. What we know is that 10 to 20 percent,
depending on what you're readings--there
are only
three or four studies--do propagate from
distal to
78
proximal.
We know about 5 percent propagate to PE.
We don't know the actual recurrence rate
of what
further DVTs would be after, say, 6
months. We do
also know that post-thrombotic syndrome
occurs in 5
percent of patients in 2 to 7 years after
total
knee replacement. Those are the real data that we
have.
Not an area well studied.
DR. BORER: Jonathan, and then Steve.
DR. SACKNER-BERNSTEIN: I noticed that in
the trials for the study flow of patients
in
several of your trials, including THRIVE,
both
EXULTs, and SPORTIF V, that there is a
number of
patients listed as being enrolled and
then a second
number of patients listed as being
randomized. And
there's very little information in either
the FDA
or the sponsor's documents about what
happened to
those patients. So I'm wondering if you could
describe it because in each of the cases
you're
looking at probably in the range of 10
percent of
the patients who are enrolled that don't
make it to
randomization.
DR. BERKOWITZ: Maybe I'll first try to
79
answer for the EXULT and THRIVE, and then
ask Dr.
Horrow for atrial fibrillation.
For the EXULT trials, patients
were
enrolled, meaning that they were seen as
an
outpatient up to a month before the
procedure, and
then would come into the hospital, and if
they had
the surgery of interest, which was
primary total
knee replacement, then would be
randomized,
assuming they went through the
eligibility
criteria.
The most common reason patients wouldn't
go from enrollment to randomization is
that either
the--there were two: one, that the surgery was
cancelled, and then the patient wasn't
rescheduled
for the procedure--excuse me, for the
study, but
did do the procedure; the other was that
with these
trials rapidly enrolling, we had many
people lined
up but then the study--we reached our
enrollment.
Those were the two major causes.
For the THRIVE study, these
were patients
who had acute events for 6 months treated
acute DVT
and then went on to a 6-month--either
placebo or to
ximelagatran 24-milligram arm, and most
of these in
80
terms of just taking a look here--I can
just show
you what we've got in terms of that. In terms of
the ones that were not randomized, there
were 123
of those patients, and most of this
turned out to
be eligibility not fulfilled or withdrawn
consent.
And that is a common thing that patients
might
think more about the study if they want
to
participate in such a long-term--and then
I could
turn it over to Dr. Horrow.
DR. HORROW: In the SPORTIF trials, the
major reason why patients were enrolled
but not
randomized was because of the failure of
an
eligibility criterion; in particular, the
major one
was the ability to achieve two electrocardiograms
demonstrating atrial fibrillation in the
manner
specified. And as a result, the principal
investigators did not enroll a number of
the
patients whom they at first thought were
good
candidates.
DR. SACKNER-BERNSTEIN: Can I just follow
up?
One quick point in follow-up. In
the patients
enrolled in SPORTIF where many of them
were coming
81
off the vitamin K antagonist, how
many--even if it
was a minority, how many of those
patients had some
sort of clinical event that led them not
to be
randomized?
DR. HORROW: I understand your interest is
in seeing what happens to the patients
who came off
of vitamin K antagonist in the enrollment
period,
did they happen to have events. I believe that we
have some data on that, although I can't
say for
sure that all of these did not
enroll. They may
have had an event after enrollment. If you'll just
give me a moment, I'll see if we can find
these
data.
Yes, thank you. Here we see the number of
patients with primary events who had an
event
within 30 days of discontinuing study
drug, and
this would be either the ximelagatran or
the
warfarin group. And this would be during the
course of the trial. As you can see, there's not
much difference between the two groups.
I think this may address the question
that
you're getting at, which is what happens
when
82
patients discontinue their anticoagulant.
DR. SACKNER-BERNSTEIN: Well, actually, I
find that reassuring, that information,
but really
what I was getting at was the impact of
the
strategy that would be proposed based on
the study,
which is you have a patient who's on
long-term
warfarin and you're going to convert them
potentially to a new agent. There's a period that
would be followed where there's a
transition, and
I'd like to know if that transition
period is a
period that could be associated with risk
as well.
DR. HORROW: I understand better. Thank
you.
Here are some data from SPORTIF III looking
at primary events within 7 days of
randomization.
There were three patients who had a
primary event
in the SPORTIF III trial within 7 days of
randomization, and, of course, the
patients taking
VKA--all patients had to stop their VKA
in order to
begin randomization. And there were two events in
the warfarin group and one in the ximelagatran
group.
DR. BORER: Steve?
83
DR. NISSEN: I want to come back to the
blinding issue again, and we've been
dancing around
it so let me just come to the point.
Something extraordinary
happened in
SPORTIF III and SPORTIF V. In SPORTIF III, I
calculate a hazard ratio of 1.39, 1.40
that's in
favor of ximelagatran. And in SPORTIF V, the
hazard ratio is 1.35 in favor of
warfarin. And so
you have almost a completely opposite
effect on the
point estimates, which, you know, is
really unusual
when you consider the similarity of the
trials.
So we're al trying--we're all
sitting here
looking at the briefing document, and
we're trying
to figure out what could possibly have
happened
here so that, you know--I mean, there's
essentially
a 39-percent greater risk for warfarin in
SPORTIF
III and a 39-percent greater risk for
ximelagatran
in SPORTIF V. And the only big difference in the
two trials is that one was blinded and
one wasn't.
And so most rational people who
look at
that would say, well, we're going to
believe the
blinded results, we're not going to
believe the
84
unblinded results. And so this is a real
credibility issue, and I think we might
as well
just put it on the table and get your
reaction to
it.
DR. HORROW: In fact, there are many
differences between SPORTIF III and
SPORTIF V that
are confounded with the open-label and
double-blind
nature of those two trials. The first and foremost
is geography, namely, that one study was
conducted
in Europe and Asia and the other in North
America,
and practice issues may pertain.
Secondly, although SPORTIF V
patients more
often had hypertension, their blood
pressures were
6 mm mercury lower, on average, than
patients in
SPORTIF III.
And, third, there was an
artificially
intense control of INRs in SPORTIF V
relative to
SPORTIF III, because in SPORTIF III there were
over
270 clinical laboratories conducting INR
measurements, but there were essentially
two in
SPORTIF V, achieving some kind of
standardization
that is difficult to quantify.
85
Another aspect that is
important to
consider is that the ximelagatran rates
were
identical in the two trials. And in the warfarin
trials--I'm sorry, in the two trials, the
ximelagatran rates were identical, about
1.6. The
warfarin rates appear disparate. But those rates
are actually within the range of rates
that are
seen in prior stroke prevention trials.
What we may be looking at here
is another
manifestation of the variability of
warfarin. This
slide shows in yellow the warfarin rates
from the
six index trials, in orange the two rates
from the
SPORTIF trials, and in dark brown the
meta analysis
rate for the trials in yellow. And as you can see,
the SPORTIF rates are within the range of
the
warfarin rates from the previous trials.
I hope that gives some
perspective.
DR. FLEMING: Could you put that slide up
again?
Can I follow up?
DR. BORER: Sure.
Let me just put some
ground rules here, though. Steve has highlighted
what will be one of the key issues for
discussion
86
later on, and rather than get into it in
great
detail here and get bogged down for the
next hour,
perhaps we can deal only with issues of
fact, and
then we'll get into the evaluation of
those facts a
little bit later.
But with that in mind, go
ahead, Tom.
DR. FLEMING: If you could put that slide
up, I just think for clarification, I
don't think
that the point you just raised really
answered
Steve's question. Steve's question had more to do
with the heterogeneity in the relative
risk
estimate across to pivotal studies. This is
getting at the heterogeneity of the
control arm
event rates across trials. And, in fact, those are
different phenomenon. This really gets at the
unreliability of noninferiority
comparisons because
of this tremendous heterogeneity, which
is a
separate issue.
While I have the mike, could I
ask a
question that I had in mind? That is, one of the
things that's always concerned me in
trials with
venograms is that we end up with a lot of
missing
87
data, far more than what this committee
would be
used to accepting in a manner to maintain
integrity
of randomization. I think you had 20 and 15
percent, respectively, missing the
outcome
assessments in EXULT A and EXULT B.
With that in mind, and also
wanting to
really focus on what are not surrogates
but true
clinical endpoints, endpoints that
reflect tangible
benefit to patients, I struggle to look
for what
are those measures that are really
tangible that
are measured uniformly in patients. Could you show
Slide CE-19 as we look at EXULT A and
B? Two of
these measures are pulmonary embolism and
death
that should be assessed, I'm assuming,
and
available in all patients. Your survival figures
here reflect, if I pool here, five deaths
against
three.
The agency on page 26 of their briefing
document has ten against four, so you're
missing
five deaths in the Exanta arm and one in
the
warfarin arm. Could you clarify that discrepancy?
DR. HORROW: If I may first address the
issue of the heterogeneity, then we can
go on to
88
the issue with EXULT.
If I'm not mistaken, you're
referring, in
terms of the heterogeneity in SPORTIF, to
what may
be called a study by treatment
interaction, the
difference in sampling and getting one
set versus
the other. And I think it's important to
understand that in each case,
noninferiority was
satisfied; that is, looking at the data
just in
those terms and how those numbers are
sorted does
not take into account the noninferiority
design of
the trials and that the success is
determined by
whether or not it meets the
noninferiority
criterion.
The heterogeneity result which
we've
looked at is not robust to sensitivity
analyses
like the primary results are robust. So, for
example, if one looks at primary events
plus
all-cause mortality, which was an
endpoint
suggested by the agency, the
heterogeneity
disappears and the p value is 0.23. And if you
look at other prespecified outcomes, such
as major
bleeding, there's no suggestion of
disparity there.
89
The heterogeneity p value is 0.81. For total
bleeding it's 0.275.
And so we view the idea of
disparate
results in the two trials with some
suspicion and
think that we need to be very careful how
we
interpret those primary results in terms
of being
disparate or the same. We view them as sampling
from the same pool and getting two
separate results
and that the best estimate of the data
comes from
pooling.
I'd like now to--
DR. FLEMING: Given that you didn't answer
my question and you provided a different
answer,
let me respond to the answer you just
gave. The
question that Steve asked is why was
there such
heterogeneity in relative risk
estimates. The
answer that you gave was there's a lot of
heterogeneity in the control arm, in the
warfarin
rates across trials. Logically, I would assume
that if you're saying when the warfarin
rate in
truth is different across trials, we
should expect
a
different treatment effect, it really makes me
90
worry about doing a noninferiority trial
where you
have to rely on historical evidence.
Could you answer, though, the
question
that I'd asked here about the discrepancy
between
your data here and the FDA briefing
document?
DR. HORROW: I'd like to ask Dr. Scott
Berkowitz, who was the medical person for
this
particular trial, to address this issue.
DR. BERKOWITZ: Yes, Scott Berkowitz,
AstraZeneca. I just wanted to say in terms of the
venography rate--I have the data to show
you, but
in terms of venograph, these two trials
had the
highest adequacy of evaluability ever
done in
clinical trials for pivotal purposes.
DR. FLEMING: That may be, and yet the
reality is we're still lacking 15 to 20
percent of
our randomization cohort, and we no
longer are
assured of integrity of
randomization. So could I
get the answer to my question?
DR. BERKOWITZ: So for what you saw, those
data that you saw in the briefing packet
were for
the overall study, so you can see it's
ten and
91
four, but I'm going to--could I have the
next
slide?--show you the breakout for
treatment, which
is the primary endpoint--
DR. FLEMING: So, in fact, what I do want
is the entire study, ten and four. So is the
clarification CE-19, then--
DR. BERKOWITZ: Could we go back?
DR. FLEMING: Then the reason CE-19 is
leaving out the five deaths and one death
is that
those occurred in the non-80, 85 percent?
DR. BERKOWITZ: The deaths--I'm sorry.
Say that again? I'm sorry.
DR. FLEMING: What is the reason that your
slide here leaves out five deaths and one
death?
DR. BERKOWITZ: That slide showed the
primary endpoint which included the
treatment
period of day 7 to 12 days as opposed to
the
overall, which showed only this study and
the next
one, if you want to see the breakout.
DR. FLEMING: Good.
And so that is--could
you show it again?
DR. BERKOWITZ: Oh, yes.
Can I see the
92
next one?
Thank you. We want to see now the
breakout between the treatment--
DR. FLEMING: So that the total deaths are
as here, they are as there in the FDA
briefing
document, ten against four in the wrong direction.
And pulmonary embolism is, according to
the FDA
briefing document, four against five as
reported by
the FDA.
DR. BERKOWITZ: Yes.
Well, you can see
down--for the treatment period, as you
can see,
there was one in the ximelagatran 36 group
and none
in warfarin. During the follow-up period, there
were four in the ximelagatran group and
zero in
warfarin.
DR. FLEMING: And so in an ITT analysis
that does include all patients and
focuses on,
among the most clinically relevant
endpoints, death
and pulmonary embolism, it appears that
there are
actually numerically an excess of events
in the
Exanta group. By my count there are 15 events
against 10 events, and that's your
numbers as well.
Is that correct?
93
DR. BERKOWITZ: Well, except that the
numbers that you're seeing in follow-up
are after
patients are off treatment but they get
seen in 4
to 6 weeks.
DR. FLEMING: I want ITT, and that's what
it looks like. Is that correct? It's 15 against
10 in the wrong direction? Just is the FDA summary
correct on page 26?
DR. BERKOWITZ: Yes.
DR. FLEMING: And one other quick
question, if I could. Again, wanting to try to
focus on an ITT of a critical endpoint,
all-cause
mortality, in THRIVE III could you show
us the ITT
summary?
This is the placebo-controlled trial
where we see a substantial efficacy
result on the
symptomatic endpoint. Could you show us the ITT of
the survival curves for that trial?
DR. BERKOWITZ: I'm not
certain--did you
want to see the slide that we showed for
the
original presentation?
DR. FLEMING: I believe it's corresponding
to the page 7, Figure 1 in your briefing
document.
94
DR. BERKOWITZ: Let me bring that up. I
just want to be sure it's the same one
that we saw.
DR. FLEMING: The one that I am in
particular looking for here, because that
figure
includes all the data from all the
trials, is in
particular THRIVE III with ITT analysis
of
mortality over the time frame that you
followed
these patients.
DR. BERKOWITZ: Yes, okay, and that's what
we were--yes, I'm sorry. So here you go. This is
the slide.
DR. FLEMING: Mortality.
DR. BERKOWITZ: Yes.
DR. FLEMING: All-cause, ITT.
DR. BERKOWITZ: I'm sorry.
I still don't
understand what you--just the mortality
slide?
DR. FLEMING: Yes, as you have in Figure 1
of your briefing document.
DR. BORER: You wanted to see only for
THRIVE, or you wanted to for the--
DR. FLEMING: Either way, if you--okay.
DR. BERKOWITZ: This is the slide in the
95
briefing document that you're speaking of
with all
the mortality.
DR. FLEMING: Okay.
And could you--so the
THRIVE is, in fact, the--
DR. BERKOWITZ: I'm sorry, yes, the THRIVE
is the lowest curve there, the
ximelagatran versus
placebo in the lowest group.
DR. FLEMING: And so essentially, while
I'm focusing on THRIVE, the evidence here
would
suggest, even in placebo-controlled
comparisons,
there's strong suggestion of no
differences in
survival.
DR. BERKOWITZ: Well, I mean, they're
lower with the ximelagatran group, but
not a strong
difference.
DR. FLEMING: I'm sorry.
I don't--the
curves look overlapping in the THRIVE
III, and in
the
other studies they are very overlapping as
well.
DR. BORER: Steve?
DR. NISSEN: Yes, I just had one more
question, still trying to probe to
understand the
96
differences between SPORTIF III and
SPORTIF V.
Could you show us the INR values, that
is, the
degree of anticoagulation control in
SPORTIF III
and SPORTIF V for the warfarin arms.
DR. BERKOWITZ: I'll ask Dr. Jay Horrow to
present that.
DR. HORROW: I'm sorry.
I missed the last
two words in--
DR. NISSEN: Yes, I just want to see in
the warfarin arm of the trials, I want to
see what
the INRs looked like in SPORTIF III and
SPORTIF V.
DR. HORROW: Okay.
I believe these data
will address your question. There were almost
100,000 different INR values, and this
summary
perhaps helps. Here we have SPORTIF III and
SPORTIF V and the percentage of time in
specific
ranges.
DR. NISSEN: It does.
DR. HORROW: Okay.
Thank you.
DR. BORER: A final question of fact,
Jonathan?
DR. SACKNER-BERNSTEIN: I think the key
97
thing is that all of the slides that show
ITT are
not true ITT analyses. It's not just THRIVE. It's
THRIVE and EXULT, and they list that in
the
briefing document. There are a lot of numbers
where those are different, so we should
just
interpret it that way.
DR. FLEMING: It was part of the reason
for my asking the question. I wanted to get a
verification that we were being shown,
for
endpoints such as mortality, a true
ITT. And I
understand that they're telling us they
are showing
us a true ITT where you have uniform
follow-up
through a given calendar date at which
the study
data freeze would have occurred, and you
would have
complete follow-up on mortality for all
patients.
Is that what that Figure 1 showed?
DR. HORROW: Yes.
DR. BORER: Okay.
Thank you, Jay.
DR. HORROW: May I introduce Dr. Sunita
Sheth, who will discuss particular
aspects of
safety for ximelagatran.
DR. SHETH: Good morning.
I'm Sunita
98
Sheth, Senior Director of Clinical
Research at
AstraZeneca.
You've just seen the efficacy
data
supporting the benefit of ximelagatran as
an oral
anticoagulant. I'll now review the clinical safety
date.
The analysis comes from a large data set
with more than 30,000 subjects, many of
the
patients involved having serious
underlying disease
and receiving multiple drug therapy.
First, I'll discuss by
indication the
adverse events and bleeding
profiles. Efficacy for
any anticoagulant is balanced by risk of
bleeding.
Indeed, bleeding and the prevention of
thrombosis
derive from the same action of drug. That's why
bleeding was a prespecified endpoint in
the pivotal
trials.
And major bleeding was adjudicated in a
blinded fashion in all Phase II and Phase
III
trials.
Then I'll focus on two specific topics:
myocardial ischemic events and the
hepatic
findings.
Finally, I'll conclude with a review of
overall mortality and summarize the key
points for
each indication.
99
It may help if I display how
we've
organized the large data set. It divides logically
into three groups: Phase I, surgical, and
nonsurgical populations. The Phase I population,
composed primarily of healthy volunteers
dosed for
up to 8 days, didn't present any safety
signals.
Surgical patients, mostly from the
orthopedic
studies with dosing up to 12 days, have
different
safety issues, in particular,
perioperative
bleeding, and so they are reviewed as a
separate
group.
The nonsurgical population
primarily
received drug for more than 35 days and
provides
the core safety evaluation of long-term
dosing,
with exposure up to 4 years. Each population pool
is
large, allowing a detailed assessment of safety
in each case.
I will first review the safety
for the
surgical indication. The North American surgical
population has been termed "the
warfarin comparison
pool" and provides the safety data
for the
indication under consideration today,
with
100
post-operative dosing of either oral
ximelagatran
or warfarin after total knee replacement
surgery.
This pool includes data from three Phase
III
trials:
the two EXULT trials as well as an earlier
study evaluating 24 milligrams versus
warfarin.
Overall, it includes 5,236 patients.
In all graphs, ximelagatran
will be shown
in a shade of orange and the comparator
in gray.
Here's the summary of adverse
events for
the surgical pool. Both treatment groups showed a
similar frequency and type of adverse
events.
There didn't appear to be any dose
response
comparing the 24- and 36-milligram
doses. We can
look more closely at the EXULT trials
where both
major and minor bleeding events underwent
independent adjudication. Rates of major bleeding,
shown at the bottom of each bar, were 1
percent or
less in all treatment groups, with no
statistically
significant differences for major
bleeding alone or
for the combination of major and minor
bleeding,
for
which respective p values are shown.
When you look at the data for
the proposed
101
36-milligram dose, there wasn't a
difference in
surgical outcome parameters, such as
wound hematoma
or intra-articular bleeding. Additionally, the
proportion of patients receiving
transfusion and
the volume of transfusion were similar in
each
group.
Now, let me turn to the nonsurgical
patients who comprise the long-term
dosing group.
This group is called the long-term
exposure or LTE
pool, with patients from all the Phase II
and Phase
III studies conducted so far involving
dosing
beyond a month's duration. In addition to patients
from the atrial fibrillation and venous
thromboembolic secondary prevention
indications,
we've included data from two other
disease areas
where significant trials have been
conducted,
patients undergoing initial 6-month
treatment for a
venous thromboembolic event and patients
post-acute
coronary syndromes. The overall ximelagatran
exposure is substantial, a total of 6,768
patient
years, with a median exposure of 370
days.
Across this population, doses
between 20
102
and 60 milligrams have been used,
although the
majority of patients, 75 percent of them,
received
36 milligrams twice daily. The comparator group
includes both placebo as well as warfarin
and is
termed "the comparators'
group." In this group, 20
percent of patients received placebo.
I'll now comment on the different
indication pools.
In the VTE extended prophylaxis
pool, both
ximelagatran and placebo groups
demonstrate similar
frequency and types of adverse
events. The
incidence of serious adverse events and
discontinuations was actually lower in the
ximelagatran group compared to placebo.
In the same group, major
bleeding occurred
rarely, affecting six patients in the
ximelagatran
group and five patients in the placebo
group.
Ximelagatran and placebo groups also did
not differ
with respect to major or minor bleeding
events.
In the atrial fibrillation
pool, the same
frequency and types of adverse events
were recorded
in both the ximelagatran and warfarin
groups.
103
Discontinuations were higher in the
ximelagatran
group, not because of symptoms but mainly
due to a
protocol-mandated discontinuation for ALT
elevation. I'll discuss this in detail shortly.
In the atrial fib population,
the rates of
major bleeding with ximelagatran did not
differ
from those with warfarin. Minor bleeding events
occurred quite often in these trials and
for that
reason did not undergo adjudication. Here we see
the event rates for patients with one or
more major
or minor bleeding events. Total bleeding occurred
significantly less often with ximelagatran
than
with warfarin, with a p value of less
than 0.001.
Overall, with regard to adverse
events and
bleeding, ximelagatran compared to
well-controlled
warfarin following total knee replacement
surgery,
compared to placebo and extended
secondary
prophylaxis of VTE, and compared to
warfarin in
atrial fibrillation patients demonstrated
no
important differences in adverse events,
bleeding
profile, or the safety profile of the 24-
and
36-milligram doses. In addition, a detailed
104
subgroup analysis for bleeding supports
the
proposed fixed-dose approach for all
types of
patients studied.
I'll now review two special
safety topics,
coronary artery disease and the hepatic
findings.
First let's address the coronary artery
disease
findings.
The agency has noted a possible
imbalance
in the frequency of myocardial
infarctions. Shown
here is Table 12 from the FDA briefing
document.
The events shown here are
investigator-reported
events.
Note that the absolute number of
myocardial infarctions observed in the EXULT
trials
was small, and there appears in a
post-hoc analysis
to be a significant difference with a p
value of
0.049.
However, this difference is driven by a
single trial, EXULT A. Furthermore, an analysis of
other coronary artery disease events
failed to
reveal any significant difference.
FDA Table 40 shows
investigator-reported
coronary adverse events from selected
trials from
the long-term pool. This analysis suggested an
105
increased frequency of total coronary
adverse
events in the VTE treatment
population. When VTE
treatment and extended prophylaxis are
evaluated I
a post-hoc pooling and analysis, the p
value is
significant for both myocardial
infarctions and
other coronary artery disease
events. However,
this finding was not observed in the much
larger
atrial fibrillation pool. In addition, the trial
in acute post-coronary syndromes where
benefit was
demonstrated is not included in this
analysis. In
fact, when all three groups are pooled,
no
significant difference is observed for
either
myocardial infarctions or other coronary
events.
In addition to the
investigator-reported
events, the SPORTIF trials in atrial
fibrillation
with an active comparator, warfarin, and
the ESTEEM
trial in the post-ACS setting versus
placebo
provided an independent and objective
assessment of
myocardial infarctions. In fact, adjudicated
events in these trials represent over 90
percent of
all MIs across the program. Here, evaluation of
the SPORTIF trials demonstrated an
identical
106
incidence while the ESTEEM trial
demonstrated an
actual reduction in myocardial
infarctions.
It is also relevant in this
context that
across the whole program we have no
evidence of any
rebound effects producing MIs after
ximelagatran
treatment was stopped.
So with regard to coronary
artery disease
adverse events, while a concern was
raised
regarding a potential imbalance in
events, a more
comprehensive analysis focusing on both
investigator-reported and objectively
assessed
events fails to identify an increased
risk.
I now want to turn to the
unexpected
results, the hepatic findings, and present a
detailed review.
We've taken the findings very
seriously,
and from the large database individual
case
analysis and consultation with hepatic
experts,
we've produced a thorough
assessment. I'll first
review the laboratory findings followed
by the
adverse event data.
Preclinical toxicology and the
Phase I
107
studies did not demonstrate any hepatic
safety
issue.
The surgical studies with up to 12 days of
dosing didn't show any hepatic changes
with
ximelagatran, just the well-recognized
enzyme
elevation seen with heparin. In the first Phase II
long-term dosing study with ximelagatran
in the
atrial fibrillation patients, a signal of
an
asymptomatic increase in ALT greater than
3 times
the upper limit of normal was noted. Therefore,
the standard laboratory testing that was
being
performed early in the development
program was
increased in the Phase III studies.
The liver function testing
panel consisted
of alanine amino transferase, or ALT;
aspartate
aminotransferase, or AST; alkaline
phosphatase, and
total bilirubin. These tests were performed
monthly for the first 6 months of
exposure, then
every 2 months up to one year, and then
quarterly.
In addition, weekly testing and
discontinuation
criteria were defined. These criteria were
strengthened after one case of
biopsy-documented
hepatic necrosis.
108
As mentioned, there was no
increase in ALT
greater than 3 times the upper limit of
normal in
ximelagatran patients undergoing total
knee
replacement compared to warfarin during
treatment.
At the 4- to 6-week follow-up, there were
eight
patients in the ximelagatran group and
three in the
warfarin group that developed an increase
in ALT.
In general, these increases occurred 3
weeks after
discontinuation of drug. It's important to note
that two patients with the transaminase
elevation
in follow-up in the ximelagatran group
had received
low-molecular-weight heparin. The ALT elevation in
all patients but one in each group is
documented as
resolved.
We believe that patients undergoing
orthopedic surgery with short-term dosing
of
ximelagatran are not at an increased risk
of ALT
elevations or liver injury.
Now, let me summarize the
incidence of
enzyme elevations of the long-term
exposure pool.
The incident of ALT greater than 3 times
the upper
limit of normal was 7.9 percent for
ximelagatran
compared with 1.2 percent for
comparators. It's of
109
interest to note that there was no
difference
between groups for isolated elevations of
bilirubin. The vast majority of these enzyme
elevations were asymptomatic.
Our experience shows that the
time
signature for ALT elevation follows a
consistent
pattern.
This graph depicts the number of patients
with first ALT greater than 3 times the
upper limit
of normal over time. The y axis represents the
cumulative risk of an ALT greater than 3
times the
upper limit of normal and the x axis time
in
months.
As can be seen, the occurrence increases
above background rates after 1 month and
approaches
background rates after 6 months. Ninety-three
percent were detected during the first 6
months,
and 98 percent within the first 12
months.
I now want to turn to the
disposition of
patients with an ALT increase. Of the 546 patients
in the ximelagatran group that had an
increase to
greater than 3 times the upper limit
normal, 46
percent of patients continued to
treatment and
completed the study. The other 54 percent
110
discontinued study drug. Overall, 96 percent of
ximelagatran-treated patients returned to
less than
or equal to 2 times the upper limit of
normal ALT,
regardless of continuation or
discontinuation of
drug.
Of the 74 patients in the comparator group,
31 percent continued treatment, and the
other 69
percent discontinued treatment. Overall, 93
percent of comparator-treated patients
recovered.
The algorithm allows
continuation of
treatment for mild and transient
increases on drug.
These data demonstrate the reversibility
of the ALT
increases.
Patients who continued drug
recovered by a
median of 28 days, and those who
discontinued drug
by a median of 40 days. Eighteen patients were
rechallenged early in the program. Only two
patients had a subsequent ALT rise. One pt with a
peak ALT of 10 times the upper limit of
normal was
rechallenged after 65 days and did not
have a
repeat elevation until 2 months
later. The second
peak was at 3 times the upper limit of
normal, and
the drug was discontinued.
111
The second patient did not have
a true
rechallenge, but had multiple episodes
above 3
times the upper limit of normal, but
overall
recovered with continuation of the
drug. There was
no evidence in these or any other
patients for an
immunoallergic response.
Hepatic experts that we
consulted
suggested that the elevation of ALT
greater than 3
times the upper limit of normal and
clinical
jaundice, in the absence of an
alternative
diagnosis, can be considered a signal of
severe
hepatic injury. We selected a more conservative
definition to standardize the levels and
timing and
included cases with ALT greater than 3
times the
upper limit of normal and bilirubin
greater than 2
times the upper limit of normal, the
latter
occurring within one month of the ALT
rise.
A total of 37 patients, or 0.53
percent,
in the ximelagatran group had this
concurrent
elevation of ALT and bilirubin, compared
with five
patients in the comparators' group, with
an
incidence of 0.08 percent.
112
Please note that one additional
case has
been included in this analysis at the
request of
the FDA.
We had fully documented this case
involving a fatal GI bleed in the
submission and
had also highlighted it as a case of
interest in
the safety review.
I'll now review the outcome in
patients
with a concurrent increase in both ALT
and
bilirubin. Confounding diagnoses were noted in 25
of the 37 patients on ximelagatran. Seven patients
in the subset died of unrelated
causes. Twelve
patients did not have an alternative
diagnosis for
the enzyme elevation. Of these 12, two died with a
GI bleeding event and will be discussed
shortly.
The ALT and bilirubin in all other
patients
recovered. Of the five cases in the comparator
group, four had an alternative diagnosis,
and only
one had an unexplained increase. Two patients died
from pancreatic cancer. The other patients
recovered.
We have been investigating a
possible
mechanism for the hepatic changes, but so
far this
113
has not been elucidated. Preclinical studies
evaluating reactive metabolites,
mitochondrial
dysfunction, and protein binding have not
been
revealing. There is no evidence for involvement of
the P450 system. The asymptomatic and
nonprogressive pattern of ALT increase
has been
noted with other drugs, including
tacrine, INH,
amiodarone, among others.
We wanted to understand if
there's a
subgroup that's at increased risk. Because the
number of patients with concomitant ALT
and
bilirubin is so low, this analysis was
performed on
the occurrence of ALT greater than 3
times the
upper limit of normal. Therefore, these results
should be interpreted with caution. A step-wise
logistic regression was performed looking
at
demographic factors, statin use, and
baseline
disease.
As expected, the most significant factor
in this analysis was ximelagatran
treatment with an
odds ration of 6.82.
Other factors that demonstrated
statistical significance all had an odds
ratio of
114
less than 2. These includes patients post-ACS,
patients being treated for an acute
venous
thromboembolic event, body mass index
less than 25
kilograms per meter squared, and female
gender.
Statins and creatinine clearance were not
identified as significant factors.
The variable of ALT greater
than 3 times
the upper limit of normal is generally
asymptomatic
and reversible. Therefore, this analysis does not
allow a prediction for those at risk for
severe
liver injury. We are, therefore, recommending ALT
testing for everyone who starts long-term
treatment
with ximelagatran.
Now let's look at the adverse
event data
from these patients. No difference is noted
between groups for clinical hepatobiliary
adverse
events.
T2B I will now briefly review
three selected cases in the group of
patients with
concomitant increase in ALT and bilirubin
associated with ximelagatran. These cases were
selected by the FDA as three deaths with
associated
115
severe liver injury. The first two cases occurred
on the first algorithm, and the third
case on the
second more conservative algorithm. The second and
third case did not demonstrate compliance
with the
algorithm in effect at the time. The deaths in all
three cases are also confounded by other
factors.
In the first two cases, the ALT
and
bilirubin increase was unexplained, and
the
terminal event in both cases was a GI
bleeding
event.
The first patient, an
80-year-old male,
had a hepatic biopsy with documented
hepatic
necrosis about 1 month before death. This patient
had evidence of decreased hepatic
function.
However, the ALT was recovering when he
died from a
perforated duodenal ulcer. This patient had been
on prednisone.
The second case presented
hypertensive to
the hospital with an elevated ALT of 11
times the
upper limit of normal and a bilirubin of
1.4 times
the upper limit of normal after missing
two weekly
tests for an elevated ALT. The INR was 3.4 and the
116
APTT was 69 seconds. His last dose of ximelagatran
had been earlier that evening. The patient had a
prior history of duodenal ulcer and
Bilroth II
anastomosis with bleeding at the site
detected on
this admission. During the 24 hours from the
admission to death, the patient received
massive
transfusions. During this time his bilirubin
increased from 1.4 times the upper limit
of normal
to 9.4 times the upper limit of normal,
with 50
percent noted as indirect bilirubin. At the time
of death, the bilirubin was 7.3 times the
upper
limit of normal and the ALT less than 2
times the
upper limit of normal.
The third case was a death due
to
fulminant reactivation hepatitis B with
an elevated
ALT upon study initiation. This patient was on two
immunosuppressive drugs: prednisone and
azathioprine. Ximelagatran was not discontinued
when the ALT reached greater than 5 times
the upper
limit of normal as recommended. The patient had a
rapid and fulminant course attributed to
the
hepatitis B. However, the investigator could not
117
rule out that the drug did not contribute
to the
fulminant course.
To summarize the hepatic findings, ALT
elevations greater than 3 times the upper
limit of
normal occurred in 7.9 percent of
ximelagatran-treated patients, occurring
primarily
within the first 6 months. The elevations were
typically asymptomatic and reversible,
without any
evidence of an immunoallergic
reaction. An
incidence of 0.5 percent of concurrent
ALT greater
than 3 times the upper limit of normal
and
bilirubin greater than 2 times the upper
limit of
normal was observed. Exposure response suggests
that exposure is not predictive of
individual risk
of transaminase elevation, and no patient
subset
was identified to be at higher risk of
developing
severe hepatic injury.
Based on the data, we are
proposing ALT
testing in the label reflecting the more
conservative testing schedule used in
clinical
trials.
To make sure that ALT testing becomes the
standard of care with ximelagatran, we
also
118
submitted a risk minimization plan which
set out
our initial proposals to support ALT
testing in
practice.
This proposal was developed after
extensive external consultation and field
testing,
but we recognize that it may need to be
developed
further in the best interests of ensuring
patient
safety.
We have a meeting arranged with FDA on
this topic in the near future.
A few comments on the
principles of our
Risk Minimization Action Plan. The ultimate goal
of the plan is to prevent any hepatic
failure
caused by treatment with
ximelagatran. To do this,
the Risk MAP will help to ensure
compliance with
labeled ALT testing recommendations. This proposal
was developed to provide access to
ximelagatran by
those patients who will benefit while
minimizing
risk.
It targets patients, physicians, and
pharmacists. It has a strong educational focus and
is enhanced with practice management
tools and
special packaging. In addition, following
discussions with FDA, AstraZeneca will be
proposing
additional enhancements to ensure our ALT
testing
119
recommendations are followed. Finally, we have
proposed continuous evaluation of program
effectiveness.
AstraZeneca understands that
the full
benefit of ximelagatran can only be
realized if it
is used in accordance with the labeled
recommendations, and
to that end we are committed to
developing the specifics of the program
in
consultation with the agency.
To complete the assessment of
safety, we
will finish with the overall mortality in
the
long-term exposure pool to get an
overview of risk.
The patient population was primarily an
elderly
population with multiple comorbidities
and
concurrent medications. Despite an increase in ALT
in the ximelagatran-treated patients, no
difference
in all-cause mortality was noted. Mortality was
similar in the ximelagatran group compared
to
patients on placebo, patients on placebo
plus
aspirin, and patients on warfarin.
Let me finish by summarizing
the
benefit/risk comments for each
indication.
120
Ximelagatran prevented venous
thromboembolism
and/or all-cause mortality compared with
warfarin
in total knee replacement surgery with a
number
needed to treat of 12. No difference was seen in
bleeding, transfusions, or surgical
outcome.
Ximelagatran demonstrated clear
benefit
over placebo with a number needed to
treat of 10 in
the long-term prevention of recurrent VTE
events.
This included a clinically important reduction
in
pulmonary embolus, a condition that can
result in
serious morbidity and mortality. The incidence of
bleeding was comparable to placebo.
Ximelagatran was as effective
as warfarin
in reducing the risk of stroke and other
thromboembolic events in patient with
atrial
fibrillation. Bleeding was lower on ximelagatran.
With regards to the hepatic findings,
while the
risk per year for stroke or venous
thromboembolism
is continuous, the risk for an ALT rise
and
subsequent severe liver injury is limited
primarily
to the first 6 months of ximelagatran
therapy. But
the protection from a thrombotic event by
121
ximelagatran is continuous and consistent
over
time.
To aid effective management of
the hepatic
risk, ALT testing will be recommended in
our
proposed labeling, and in addition, we
have
submitted a Risk Minimization Action Plan
which we
will discuss further with the FDA.
We conclude that ximelagatran,
the first
new oral anticoagulant in over 50 years,
does have
a positive benefit/risk in each proposed
indication
provided that the drug is used
properly. We look
forward to your comments and further
dialogue with
the agency.
Thank you. I'll take questions.
DR. BORER: Dr. Sheth, I think we need to
take a break. I've been chastised when we haven't
done that. So we'll take a 10-minute break, and
then we'll go on to the questions of fact
about the
safety data, and I think we can then go
on to
Jonathan Halperin's presentation, and
we'll just
make up the remaining FDA time after the
public
comments later so that you can get your
whole
122
presentation in.
So we'll take a 10-minute break
right now.
Look at your watch because 10 minutes
from now
we're going to start again.
[Recess.]
DR. BORER: While everybody is getting
back in here and sitting down--or not
sitting
down--let me raise an issue for you to
begin to
think about as people are coming back in.
Steve Nissen asked earlier
about
pharmacological evidence of rebound, and
there
didn't appear to be significant rebound,
although I
don't know what that means in the context
of
studies with limited power. But there didn't seem
to be significant rebound of
pharmacological
effects, although the follow-up, as I
recall, was
relatively short. So we don't know about late
pharmacological changes. But as I look at these
data from each of the trials, I'm struck
with a
difference between the on-treatment and
post-treatment frequency of major adverse
cardiovascular events that I'd like to
hear some
123
discussion about from you. Is this real or is it
not?
That is that if you look at the number of
myocardial infarctions or other cardiac
events that
occurred on ximelagatran versus the
comparator, the
numbers were different but not all that
different.
It depended on the trial. It varied from trial to
trial, and we can talk about that
potential adverse
event disparity later. But I'm concerned or I want
to ask about something else.
If you look at the number of
events that
occurred on-treatment, the numbers were
relatively
close one way or the other from trial to
trial to
trial.
If you look at the numbers that occurred
post-treatment, the proportion of
patients who had
events on ximelagatran in the
post-treatment period
was greater as a percentage of the whole
than was
the case for any of the comparators. The
post-treatment events on warfarin or on
placebo
were fewer as a proportion of the whole
of the
total number of events in those
comparator groups
than was the case with ximelagatran, and
in some
cases the post-treatment events were more
frequent
124
than the on-treatment events with
ximelagatran.
That's an observation.
Have you noted that? And is that true?
And do you have anything to say about it?
DR. SHETH: The numbers differ a little
bit between the different patient
groups. So let
me start first with the long-term
exposure pool and
some of the specific populations within
that pool.
If we can take a look again at
the--and
we're talking coronary events, Dr. Borer?
DR. BORER: Yes, we can limit it to
coronary events, however they're defined.
DR. SHETH: What you see is that, you're
right, there is a difference--I'm
sorry. Let's put
that up.
You do see an increase--and these are
both myocardial infarctions and total
other
coronary artery disease events other than
MI
compared in the VTE treatment and the VTE
extended
prophylaxis compared to warfarin. But these
numbers are actually quite small. We're talking
about a total of 16 patients here, 3, 16,
10, et
cetera, versus 1, 0, 12, 3.
125
If we take a look at two
populations where
you might say that the risk is actually
increased,
the atrial fib group had higher incidence
of both
diabetes, hypertension, for example. You don't see
that--those events, again, plus that in
the
post-acute coronary syndrome population,
which is
certainly a high-risk group for
events. Can we
take a look at the next slide? I'll come back to
the after-treatment in a second.
Ninety percent of the MIs--and
this was
during the trials--occurred in these two
settings,
and you don't see a difference there, and
you see a
benefit on treatment with ximelagatran.
But if we take a look, let's
say, at the
post-acute coronary syndrome population,
again, a
higher-risk group, after treatment
stopped, the
incidence between those two groups was
about 1.5
percent--I think it's about 1.5 percent in
both
groups.
DR. BORER: Okay.
I'm sure you're right,
and the data you just showed I think are
very
reassuring, and I think we all saw them
in the book
126
here.
But, again, I'm making a slightly different
point, and maybe the data aren't
available or
aren't sufficient to draw a firm
conclusion about
them.
What I am talking about is the
proportion
of coronary events that occurred after
stopping
treatment on ximelagatran as a percentage
of the
total number of events compared with the
portion
that occurred after stopping treatment
with
warfarin or placebo as a percentage of the total
number of events in those groups. I believe that
the proportion of events that occur
post-treatment
is higher in the ximelagatran groups
across all the
trials, if you look at trial after trial,
than is
the case for the comparators, which
raises some
question about the possibility of a
rebound
phenomenon or something else, some other
pathophysiological process that's being
allowed to
happen or occurring because of the use of
the drug
once it's stopped.
DR. SHETH: I understand what you're
asking.
We don't have that specific analysis, so I
127
won't be able to address it at this
moment. You're
asking for those proportions of patients
after they
stop treatment over the total number of
events, and
right now I don't have that analysis.
DR. BORER: Okay.
You can pull it
together later, but it's in the
books. If you look
at the data that are presented, if you
look at the
numbers, that sort of jumps out at
you. So you may
want to look at that, and you can talk
about it
after lunch or something.
DR. SHETH: Okay.
DR. BORER: Okay.
Well, why don't we go
on and see--Alan?
DR. HIRSCH: Well, just one comment to
follow up Jeff, and if you're able to
provide that
after lunch, I specifically would ask you provide
that not in the ACS population, because
the
population that will be exposed to this
if this
drug comes to market that really is
vulnerable that
I'm concerned about is that non-ACS
population.
DR. SHETH: Okay.
DR. HIRSCH: I don't want that to be a
128
Band-aid for a potential adverse effect.
DR. BORER: Steve and then Bill.
DR. NISSEN: Just so you understand what
we're concerned about--and several of us
have made
this observation--it is that because
ximelagatran
is a short-acting agent compared to
Coumadin, our
worry is that when you stop the drug,
there's some
phenomenon that goes on for a few days or
a few
weeks in which a patient has increased
vulnerability and
that that is the explanation for the
excess cardiovascular events. And we want to
understand whether you have some response
to that
that we can factor into our thinking.
DR. SHETH: Can I ask, would it help the
committee to take a look at other
thrombotic events
in terms of incident or rebound
phenomena? Because
certainly patients who are usually at
risk for
venous events might typically get those
kind of
events.
Would that help--
DR. NISSEN: It only helps a little bit.
The problem is that the pathophysiology
of arterial
and venous events are different.
129
DR. SHETH: Right.
DR. NISSEN: And so, you know, it appears
that there is this excess of arterial
thrombotic
events post-treatment, and we're trying
to
understand that in order to factor that
into the
thinking here of the committee.
DR. SHETH: Right, although in maybe the
treatment and prevention groups--
DR. NISSEN: Yes, yes.
I have another
question, and forgive me for this, but I
have to
probe on something that I think is
important. If
you could put up Slide No. CE-19,
please? I see
these patients that are going to have
knee
replacement all the time in
consultation. They
almost all get sent for cardiac clearance
because
they're older and they have a lot of
cardiovascular
risk factors, and I'll bet the other
cardiologists
at this table, like Jeff probably sees
plenty of
these as well. And so when I see them, there are
three things that I worry about. I worry about, of
course, them dying. I worry about them having a
pulmonary embolus. And I worry about them having a
130
myocardial infarction.
And so, you know, to do the
simple math
here, which is what all of us are kind of
looking
at, if you look at the serious endpoints,
the
feared complications, what you see is--in
EXULT A
and B, you see three plus six is nine
events with
ximelagatran, and you see eight events
here, PE or
death, with warfarin.
If you now put up Slide No.--
DR. FLEMING: Steve, just before you go,
those nine and eight are 15 and 10 in the
FDA
briefing document. It's worse than this. It's 15
and 10.
DR. NISSEN:
Okay. I'm trying to be--you
know, not make this any more painful than
it has to
be.
Now let's look at Slide CS-14,
so we'll
take nine and eight, so CS-14, and now I
look at
myocardial infarction, and it's 16 to
4. And so
when you put it together, you know, you
see that
the really serious events, the bad things
that can
happen to that patient I'm seeing in
consultation
131
look a lot worse on ximelagatran than
warfarin.
And so one has to ask the question: Does it really
look as good as it looks?
And so what are your thoughts
about this?
I mean, MI is as bad an outcome as PE,
isn't it?
DR. SHETH: Yes, it is.
In considering
those numbers, I won't dispute how--the
numbers
that we just looked at, they are higher
in the
ximelagatran group compared to the
warfarin group
in the orthopedic surgery
population. The only
comment I'd make is that,
unfortunately--those are
really small numbers, and the question
is: Is this
a really--a true difference? And I would
anticipate that if it was a true effect that
we
would really see a significant effect in
the
long-term group just because it's so much
larger.
We also have another study that
we started
to do in extended prophylaxis in
orthopedic
surgery, so we'll be able to collect more
data in
that study as well. But, again, the numbers are
small so it's hard to know if this is a
true
difference or not.
132
DR. NISSEN: But, of course, the
difference in the long-term studies is
that this is
one where you get the short-term
administration,
then you withdraw the drug, and so it
speaks more
to this question of an acute rebound sort
of
phenomenon.
I mean, I hope you can appreciate why
it's something that really struck many of
us on the
committee as being a problem.
DR. BORER: It's also a potentially
remediable problem, so it's important
that you
should know about it.
Bill, and then Beverly.
DR. HIATT: Yes, just to follow up on
that, it does seem like the surgical
population may
not be the same as the long-term
treatment
population, and the concept of risk
occurring--excess risk occurring in that
population
is very real.
Then the other question I have
is, turning
to the SPORTIF IV data, you didn't
present that in
any of your safety data. Is that correct?
DR. SHETH: The SPORTIF II and IV data are
133
actually pooled in the atrial
fibrillation pool
that we performed. So it included that Phase II
trial, yes.
DR. HIATT: If you look at page 97 of the
briefing document, there are several
phases to
SPORTIF IV, and I count a total of 17
deaths on
treatment versus warfarin is four. So you're
saying those deaths are included in the
overall
safety data you presented?
DR. SHETH: They are included, but I'll
just point out that there are about 2 to
3 times
more patients on--3 times more patients
on
ximelagatran than on warfarin in SPORTIF
IV. So
they're not balanced groups. The denominators are
not balanced.
DR. HIATT: Correct.
DR. SHETH: But those deaths are included
in the atrial fib pool and consequently in
the
long-term exposure pool.
DR. HIATT: Okay.
DR. BORER: Beverly, and then Dr. Sjogren.
DR. LORELL: To follow up on this concept
134
of potential rebound--
DR. SHETH: Can you speak louder?
DR. LORELL: Yes, I can.
To follow up on
the issues that were raised about
potential rebound
in the post-surgical population, can you
enlighten
us as to how investigators were
instructed to use
or no instructions on aspirin? Was aspirin
deliberately not used in that surgical
population?
And then were there any instructions at
the
termination of treatment?
DR. SHETH: Let me ask Dr. Berkowitz, who
was the physician for those studies, to
describe
the use of aspirin instructions for the
surgical
population.
DR. BERKOWITZ: Scott Berkowitz,
AstraZeneca. I didn't get the second part. The
first part was that aspirin was
precluded, kept to
a minimum, and patients weren't to be on
it
routinely.
DR. LORELL: So I think the second part,
was there a strategy in that trial when
the study
drug was stopped about reinstatement of
aspirin in
135
patients who had risk factors? You know, the point
that Dr. Borer made, this is a group with
rich risk
factors.
DR. BERKOWITZ: I'm sorry.
I think I got
all your question. The studies were designed to
leave to the discretion of the
investigators to put
the patients back on the medicine, so we did
not
prespecify how to do that.
DR. LORELL: Okay.
And related to that,
have you done any studies after
withdrawal of the
drug to look at what happens to platelet
function?
DR. BERKOWITZ: In our clinical trials for
VTE and orthopedic surgery and I believe
in the
atrial fibrillation trials, we did not do
any
platelet studies.
DR. SHETH: I can mention we actually
looked and did an analysis of patients
both on
ximelagatran, on aspirin and off aspirin,
for
events in the atrial fibrillation
pool. If you're
interested, we can show that if that
would be
helpful.
And this is not exactly the same as the
patients who discontinued after
orthopedic surgery.
136
But if you're concerned about any
increased
beneficial effect--let's see. Actually, what you
see is that there is an incremental
benefit in
patients who are on aspirin, but you see
that same
benefit on warfarin, and you don't see a
difference
of the effect between the two
anticoagulants when
aspirin is added.
DR. BORER: The question that Beverly is
asking, though, is what about after
you've stopped
the ximelagatran and the warfarin. In that period,
were people still on aspirin or were they
not? And
did the fact that they were or weren't
have any
impact on the post-treatment events?
DR. SHETH: We didn't make specific
recommendations after the trial. They were to go
on their regular medications per their
physician.
DR. BORER: Okay.
Dr. Sjogren, then Alan,
then Ron.
DR. SJOGREN: My question pertains to the
potential hepatic toxicity, and I have a
couple of
questions. One is you are proposing to follow up
patients with ALTs, and then if they go
over 2
137
times the upper limit of normal, to
follow up a
little more closely and eventually
discontinue the
drug.
I'd like to know what kind of information do
you have in the patients that you
followed up that
developed the ALT abnormality to back up
that kind
of recommendation. That's one question.
The second question is: Do you have any
information on patients with chronic
liver disease
that are treated with this medication? What
happens to them?
And one request. Do you have slides of
the liver biopsy that was done that we
can look at?
DR. BORER: Before you begin to answer,
let me just state a rule here. We're not asking
you to tell us your algorithm for
following
patients.
Dr. Sjogren is just asking about the
data that might be used to inform the
development
of such an algorithm and then the issue
of the
chronic use and the slides.
DR. SHETH: Okay.
Let me answer the
latter two questions first, and I'm going
to have
to ask for a clarification on that first
one.
138
We do not have a slide of the
hepatic
biopsy right now. The chronic disease, because we
identified early in the Phase II trial,
SPORTIF II,
that there was this asymptomatic
transaminase
increase, we actually excluded patients
who had
known hepatic disease from the trials, as
well as
patients who had an ALT above 2 times the
upper
limit of normal. So that to the best of our
knowledge, patients who--with the
exception of the
reactivation hepatitis B, should not
have, in fact,
been enrolled in the trial, and we would,
in fact,
propose a contraindication for those
patients. So
we don't have data to understand the
safety in that
group.
In terms of follow-up, are you
asking
me--you want to know what did we do to
follow up
all patients who had an elevation or what
their
outcome was?
DR. SJOGREN: No.
What I'm asking is,
you're proposing how to follow these
patients, and
you obviously during the studies have
shown us that
some patients continue on therapy despite
ALT
139
elevation. I want to know what happened to those
patients, how high those ALTs go, and
when did they
come back to normal. Did they relapse? What
happened to those patients long term to
give me an
idea that indeed, you know, your risk
management
assessment is correct?
DR. SHETH: Okay.
If I can take a look at
the scatter graphs, please? And I believe in your
briefing document you have the same
figures, and
these are actually the individual patient
data for
both patients who continued and patients
who
discontinued study drug. And I think that
individual data would be the best look,
and if I
can take you to the far--I guess my
right, is that
right?
Okay. Continued study drug here,
not real
large.
So this is the continuation of study drug,
and what you see is that patients who
continued
study drug on average returned back down
to less
than 2 times the upper limit of normal
with a
median, I believe I said, of 40 days and
a little
bit shorter if they discontinued study
drug.
I want to point out that these
scales are
140
not identical, and I apologize for
that. But if
you take a look in your briefing
document, you can
take a closer look at this.
So this is pretty much what
happened.
There was an early occurrence and a
reversibility.
I just want to make the point that those
patients
who were allowed to continue once we had
identified
and put into place some criteria were not
necessarily the patients with the highest
elevations or persistent, with the
exception of a
couple of peaks that got--that you saw up
there.
DR. SJOGREN: In the earlier slide, you
showed us that 96 percent recover. So should I
understand that 4 percent didn't? What happened to
that 4 percent?
DR. SHETH:
We can take a look at that 4
percent.
Of that 4 percent--if we can take a look
at the actual outcomes of the 4 percent
of
patients.
It was actually a total of 22 patients.
Some of those patients didn't recover because
they
died of other causes, and I will show you
what
those were. And then there are about 11 patients
141
who didn't recover for varying reasons,
and I'll
actually go through them. So that 4 percent
is--okay.
It should be--I think it's actually
supposed to be 22. So 22 and 5 comparator
patients, and if we look at the
ximelagatran
patients, we had 11 fatal cases, which
I'll review
shortly, and four nonfatal cases. Three patients,
the investigator suggested, had
alcohol-related
problems, not just routine alcohol use,
and one
patient with hepatitis C. Five patients, which we
are calling here lost to follow-up in the
sense
that in our database we did not have a
last value
that was less than 2, but we did get
follow-up from
the investigator, with one patient,
quote-unquote,
doing well, according to the
investigator. Our
goal was to find out were they alive, did
they have
hepatic disease; one who was sent on
active
military duty, so assume that person is
probably
well; one with normalized AST; the ALT
was not
done.
But we found that in our studies the AST
tracked with the ALT and typically did
not rise as
high.
And then only two that we really truly had
142
no further information.
If we could take a look at the
11 fatal
cases?
Thank you. These are what the
causes of
death were. We had at least three patients here
who had malignancy, malignancy here,
metastatic
cancer, and pancreatic neoplasia, and
then assorted
other cases, stroke, ischemic,
cardiomyopathy,
heart failure, renal failure, sudden
death, and
heart attack.
I want to just point out that
these two GI
bleeding events and this reactivation
hepatitis I
did
present earlier in detail during the core
presentation. So this is the outcome of those
patients who did not recover.
DR. BORER: Alan, and then Ron.
DR. HIRSCH: Well, actually, Dr. Sjogren's
comments were very similar to my
questions, but
maybe I'll add some speculation for the
sake of
discussion at this moment. What makes this meeting
so incredibly interesting is the fact
that we're
talking about changing thrombin as a
major
regulatory molecule, and I don't think
we've ever
143
had long-term data to understand how that
major
impact could affect human health.
So just interpreting these
data, my
concerns are, if I can go to Slide CS-23,
the
recovery time after an ALT increase one
more time.
This confused me, and maybe you've
clarified this.
It appears from this that it took a mean
of 40 days
to recover if I stopped the drug if I
were the
treating doctor but less if I continued
the drug?
DR. SHETH: Correct.
DR. HIRSCH: And that seems discordant
with the scatter graph you just showed me.
DR. SHETH: The scatter graph had two
different scales. We couldn't quite get those
scales right. The reason this is, it's probably
likely due to the recommendations we made
to
investigators about how to actually
manage ALT.
The patients who had persistent increases
or
particularly high increases were told to
discontinue, and that might indicate a
different
level maybe of severity. Those patients who had
either lower or mild increases--and I
believe you
144
have the algorithm in the briefing
document. We
can review it if it's helpful. But those patients
with lower increases or not persistent
were allowed
to continue. So they're almost pre-selected to
have different rates of recovery.
DR. HIRSCH: Okay.
That's satisfying. So
maybe we'll take a mean of 30 days here
or 35 days
for both groups. Can I just make one comment then?
There's obviously for a drug with a short
half-life
a longer-term tissue effect, which is
something
that Bev mentioned vis-a-vis platelet
function, I'm
mentioning for what happens to the liver
or the
gut.
And so any monitoring plan would have to take
into account a relatively long period of
vulnerability whereby changing drug
discontinuation
may not actually affect the natural
history of
recovery.
Does that make sense?
DR. SHETH: I'd actually like to ask Dr.
Lewis to address that question. Dr. Lewis is our
hepatologist from Georgetown
University. He might
be better able to address that.
DR. LEWIS: Thank you.
James Lewis. I'm
145
the Director of Hepatology at Georgetown.
What you saw in the other slide
in terms
of recovery, one of the other
explanations is that
for those who discontinued drug, they
had--in some
of those cases, the enzymes were higher,
somebody
was more nervous, and it may have just
taken a
little bit extra time for those values to
normalize. And that's not atypical of what we see
with many other drugs. The ones who were able to
continue drug, they were at less elevated
numbers,
and they came down.
The whole question of whether
there's any
delayed toxicity, we have not seen
that. There are
no cases of chronic injury. This is a short-term
drug.
While you're on it, you've seen the
percentage of patients whose enzymes can
rise from
mechanisms that are not defined. We don't know
what those are yet. It would be nice if we did.
We think that with the monitoring that
was put in
place--which was fairly
conservative. If your
enzymes started out normal and went to
twice normal
the ALT, you had to have your enzymes
monitored
146
much more frequently. And if they continued to
rise, the drug was discontinued, which is
what we
do with several other agents.
So if you're asking is there
going to be a
long-term effect, I don't think so. I know that
the FDA raised the issue in the
short-term
treatment, and there the numbers were
normal at the
beginning, they were normal at the end of
therapy,
and then three weeks later, as Dr. Sheth
told you,
something made those enzymes go up again,
but in a
very small group of post-op patients,
lots of other
things can happen.
DR. HIRSCH: That's reassuring. There's
no longer-term toxicity. There's still a period of
clinical vulnerability. That's different.
DR. BORER: Before you sit down, Dr.
Lewis, just one more word to complete the
answer to
that question. You said there was no evidence of
long-term problems in the people who were
treated
short term. But my recollection from the data is
that the follow-up was 4 to 6 weeks. Do we have
any information of vital status or
anything else
147
from 6 months later or a year later so we
can be
reassured that there really isn't
anything that
happened to them later?
DR. SHETH: I'll take that, actually. No,
we don't have information out to 6
months.
DR. BORER: Okay.
Tom--I'm sorry, Ron,
Tom, and Beverly.
DR. PORTMAN: As the renal part of the
cardio/renal group, I have a couple of questions.
First, in looking back at the
myocardial
infarction issue, it's noted in our
briefing book
that the major cardiovascular risks
between the
treatment and comparator group were not
different,
but not listed there is the prevalence of
chronic
kidney disease between the groups. Did you look at
that?
DR. SHETH: We did not specifically look
at that.
What we have is a breakdown of
demographics by mild, moderate, and
severe renal
impairment. But I don't think that's what you're
asking.
DR. PORTMAN: Right.
For the patients who
148
had the myocardial infarction, was their
renal
function worse than the group that didn't
have it?
DR. SHETH: Okay.
I don't have that right
at this moment, but we can take a look
into that.
DR. PORTMAN: We know that it's a very
powerful risk factor for cardiovascular
disease,
and it would be interesting to know if
there was a
difference between those groups.
DR. SHETH: And you're looking
specifically for the patients who had an
event as
opposed to those who did not?
DR. PORTMAN: Well, I think it needs to be
looked at overall, but in particular for
those who
had an event, yes. If you could provide that
information at a later point, that would
be very
interesting.
Another issue relates to the
fact that I
think the company already admitted that
they have a
lot of work to do in determining how to
dose this
drug in patients who have chronic kidney
disease.
And, in fact, taking a fixed dose of a drug in a
patient, a drug that's cleared by
glomerular
149
filtration is a little bit of a foreign
concept in
nephrology. And it's been demonstrated that there
are higher levels as the GFR decreases
and that
there is a trend toward increased
bleeding as GFR
decreases as well.
Is there a relationship or
could you
review the data for me, is there a
relationship
between the hepatic toxicity and the
decrease in
GFR?
DR. SHETH: Yes, we can take a look at
that.
Let's take a look at creatinine clearance
and how about ALT greater than 3 times
the upper
limit of normal, we can take a look at
that. And
what you see--let me wait until we get
the slide
and we'll look at the data. Here we go.
This is in the long-term
exposure pool,
and just to orient everybody, creatinine
clearance,
these patients with normal creatinine
clearance,
mild renal impairment, moderate, and
severe. This
was an exclusion within the trial, but a
few people
did sneak in. So we need to present that data
here.
150
We actually looked at it for
both ALT
greater than 3 times the upper limit of
normal,
greater than 10 times the upper limit of
normal,
and ALT greater than 3, and bilirubin
greater than
2 times the upper limit of normal.
For the ALT greater than 3
times the upper
limit of normal group, you do see a
slight increase
from 7 percent, to 8 percent, to 10 percent
in
terms of patients who develop a
transaminase
elevation. If we come to those patients with
severe hepatic injury, we don't see a
difference,
but the numbers are very small. Again, the
denominator is also very small. So it may be
difficult to see.
We do see a slight increase,
ALT greater
than 10, going from 1.8--not really much
difference
between normal and mild renal impairment,
but to
the moderate group up to 2.6 percent.
DR. PORTMAN: Well, that certainly speaks
to the need for drug dosing by GFR.
DR. BORER: Tom, and then Beverly.
DR. PICKERING: Yes, I wanted to ask you
151
about the predictors of
hetaptotoxicity. In your
slide--I think it was CS-27--you don't
list statins
as a predictor, but in the FDA analysis
they did.
And since statins also affect liver
enzymes, I
wonder if you could comment on that.
DR. SHETH: Yes, in the original analysis,
we had good treatment as a factor, and
when we put
in ximelagatran treatment as a factor,
statins
actually was not significant. So we did not
demonstrate a significant odds ratio for
statin use
and the risk of developing an ALT greater
than 3
times the upper limit of normal.
DR. BORER: Beverly?
DR. LORELL: Thank you.
That was a good
segue to my first question. If a patient did get
an elevation of LFT in the studies, were
statins
simultaneously stopped?
DR. SHETH: No.
DR. LORELL: They were not?
DR. SHETH: No.
DR. LORELL: All right.
That's helpful to
understand.
152
The other thing I'd love to
hear a comment
on, thinking about how to use this drug
in the real
world, and perhaps Dr. Lewis could
comment on this
as well, first, did you at baseline, in
your
baseline laboratory testing for chronic
use,
measure serology for hepatitis B and
C? And,
secondly, there is a little wisp of a
signal here
that could be a nothing or could be a
something,
that if you have chronic infection with
hepatitis B
or C, there could be an interaction or
synergy. So
what do we know about that, and how do we
think
about that in the real world going
forward where
both are fairly common?
DR. SHETH: Before I bring Dr. Lewis up to
answer the latter part of that question,
I just
want to say that at baseline, viral testing
for
hepatitis A, B, or C was not
performed. But once
we identified that we did have this
signal, we
actually made an effort in every patient
who had an
elevation above a certain level to
request the
investigator to do not only hepatitis A,
B, and C,
but also CMV, EBB, do an ultrasound, and
assorted
153
tests for collagen vascular
disorder. And part of
that was to understand also were there
other things
at play.
In some of those cases, some of
those
patients did have a diagnosis of
hepatitis B, not
necessarily active. But I'd like to ask Dr. Lewis
to discuss the implications of that.
T3A DR. LEWIS: The use of any
drug in somebody with chronic liver
disease, where
you're dealing with a potentially
hepatotoxic
agent, is something we wrestle with all
the time.
There are not enough patients in this
database to
know if it's a true risk factor. There are
examples, real use among patients with
HIV and
tuberculosis who are being treated with
those drugs
who have underlying hepatitis B or C,
where we know
there's a higher risk of hepatic injury in
those
settings.
Those are about the only examples that I
know where there may be a virus
interaction with
certain medications.
Whether or not we've identified
such a
thing here, the database doesn't permit
that. The
154
one patient who had the reactivation of
hepatitis B
was being treated with
immunosuppressives, had
active hepatitis B before he even got
into the
trial.
And as you heard, such patients would not
normally be prescribed the drug, and the
labeling
would contraindicate that until we do
learn more.
For patients who have cirrhosis
but not
decompensated disease, the
pharmacokinetic pattern
of the drug shows that there is a slight
rise in
the melagatran levels. But it's acceptable and it
doesn't seem to make a huge
difference. But,
again, those are patients who we wouldn't
enter
into the trials to begin with.
So there's a little bit of
information on
chronic liver disease, but not a lot.
DR. LORELL: So to press you just a little
bit on this--and you may not know the
answer to
this and are still thinking about this
with the
FDA.
Would you recommend that common hepatitis
serologies be done before starting this
drug, even
if the ALT was normal?
DR. LEWIS: No, I wouldn't.
155
DR. LORELL: You would not.
DR. LEWIS: And even though we know that
patients with hepatitis C, for example,
can have
normal liver enzymes and have mild liver
disease,
the algorithm that will be put in place
is pretty
rigorous.
And if your enzymes start to rise or if
you develop any symptoms, a full
investigation
would be done at that time.
So I don't think it would be
cost-effective to do that when you
already have
normal baseline values. I don't think we need to
know, unless there's extenuating
circumstances, if
you're pregnant or something like that.
DR. LORELL: Thank you.
DR. BORER: Steve?
DR. NISSEN: What rate of acute liver
failure is the company's point of view
that we
should assume in our deliberations about
risk/benefit? Obviously, when it comes down--and I
need to know what your thoughts are about
this.
The agency's documents suggest were based
upon Hy's
rule that we should assume a rate of 1
per 2,000
156
for acute liver failure. The observed rate is
about 1 per 2,000. What is the company's position
on the likely rate of acute liver failure
if this
drug is used in a large population?
DR. SHETH: Okay, two points on that one.
I think the first important point is that
the
number of patients which severe liver
injury
associated with death are actually quite
small. So
1 in 2,000 might represent the highest
frequency,
but, in fact, we don't know what the true
frequency
is based on the very low number. Our goal is
actually to prevent any hepatic
injury. In other
words, we do not want to market this drug
with the
possibility that patients are going to be
at high
risk for hepatic injury. And based on that, we
actually proactively submitted a risk
minimization
plan, which will be modified to assure
that
patients are protected from that risk.
So it would be difficult to hypothesize
what that would be, but our goal would be
really to
prevent it.
DR. NISSEN: I understand what your goal
157
is, and we all have the same goal
here. We really
want the same thing. You know, we want to get--if
a drug is effective and safe, we want to
get it out
there.
But we've got to decide.
So I have an agency document
that says
that I should assume a rate is 1 in
2,000. Do you
disagree with that? Do you think that the rate is
not--could not possibly be that high?
DR. SHETH: Yeah, I guess there's a little
bit of difficulty around some of the
confounding
factors that resulted in death, and not
all three
cases had evidence of acute hepatic
failure,
although I'll leave that part of the
discussion to
Dr. Lewis, if we wanted to go there.
DR. LEWIS: The true incidence of hepatic
failure in this data set, it's hard to
determine,
and there are definitions that we have to
agree to.
What's acute liver failure as opposed to
an ALT
going up and having an associated bilirubin
elevation? And Hy Zimmerman's rule, and all of us
who deal with liver disease in this room
are
students of Hy Zimmerman University, and
we all
158
know what his descriptions were based on
retrospective studies in older drugs,
that if you
had clinical jaundice and severe
hepatocellular
injury, that determined--there was an
associated
10-percent mortality. In an era where we didn't
monitor patients, this was the natural
history of
drug-induced liver disease.
We're now trying to wrestle
with what the
true signal should be based on these
kinds of
enzymes.
In almost all these patients, they're
asymptomatic. There was one individual where the
algorithm was changed for this individual
because
when his values went up above twice
normal, he
wasn't brought back, and they continued
to rise,
and the action--I didn't take care of
this
individual, and so we may never
know. But if you
look at what happened to him, he was
on--he was
recovering, but because he had
asymptomatic very
high elevations--they actually did a
liver biopsy.
They saw subhepatic necrosis. They put him on
steroids, which may have contributed to a
large
ulcer that eventually killed him because
it
159
perforated and he had a massive
bleed. But all the
indications were he did not have
encephalopathy.
His coagulation parameters off of any
anticoagulants were
improving.
I acknowledge he was
drug-related, but
would he have recovered? We may never know. This
is the actual data on this
individual. Things were
getting better, but then he died suddenly
at home
of the bleeding.
The hepatitis B patient who
reactivated,
again, may have had an increase go up
and--but for
other reasons, I think he died of
fulminant
hepatitis B. Again, he shouldn't have been in the
trial perhaps.
And the third patient we know
even less
about because he was massively transfused. He had
about 60 units of blood products. His bilirubin
rose, and his bilirubin when he was
admitted was
just over 1, and then it jumped
dramatically within
24 hours, which I think being mostly
indirect
bilirubin was related to hemolysis that
was going
on.
His haptoglobin was zero. And the
indirect
160
bilirubin--and that puts him in the
category of an
ALT that's elevated with a high
bilirubin.
Bilirubins were only total in this study,
so we're
overestimating the number of people who
could have
had indirect causes, like Gilbert
syndrome.
So if we take the worst case,
we've got 1
in 2,300 from the data set, and these
were pretty
much outside of the algorithm that's now
being
proposed, and I hope that we will be able
to reduce
that down to zero. As you said, that's everybody
hope, and it's how hopefully we can work
with
everyone here to ensure that this drug is
used
safely.
DR. NISSEN: But you understand the
dilemma that we're in here. We're going to be
asked some questions later about
risk/benefit, and
I really need to come up with in my own
mind what I
think the actual likely rate of acute
liver failure
is for this drug. And I'd like to ask you a
related question.
In clinical use, as opposed to
in clinical
trial use, do you think that the
following of the
161
algorithm is likely to be as good in
clinical
practice as it is in the trial, better
than or
worse than?
DR. LEWIS:
I hope it will be better than.
DR. NISSEN: I understand what you hope,
but I'm not asking about what you
hope. I mean,
I've got to be--I've got to pin you down
on this
because I'm going to have to vote on this
later.
DR. LEWIS: It will be better, and the
cases that are highly confounded that
we're dealing
with--and if you asked me my rate of
liver failure,
it would say it's one case, the one with
the
biopsy, because we know they had hepatic
necrosis,
and whether he died of that or not,
different
story.
But I'm going to let Sunita give you some
more insight into the monitor.
DR. BORER: Let me call a halt to that, if
I
may, and forestall that discussion. I
think that
we don't know and, in fact, we're not
going to
determine here what kind of risk
management plan
you're going to develop. The issue will be if you
could develop one, is there something
here that is
162
worth giving to patients. But that's going to come
up in the questions.
Let me limit this part of the
discussion
to one more question because John
Teerlink has been
waiting for a while, and, Dr. Sjogren,
we'll hold
yours until after Jon Halperin's
presentation
because we're going to run out of time
and we do
have other question time after the formal
presentations. And we're limited by a schedule
determination that says between 1:00 and
2:00 we
must have the open public hearing.
So let's go on to John Teerlink
now.
DR. TEERLINK: This question is actually a
point of clarification in regards to
Slide CS-27,
our old friend. I noticed that--is this slide
including all of the patients and then
for all the
risk factors, and do you have an similar
analysis
for just looking at the ximelagatran
patients? Do
you have predictors of ALT greater than 3
times
upper limit of normal for solely the
ximelagatran-treated patients, a
multivariate
analysis of that?
163
DR. SHETH: No, we do not.
DR. TEERLINK: That might be helpful for
us in deciding what are the predictors of
that
increase in ALT.
DR. BORER: Okay.
Dr. Teerlink, I've
learned that this is
"ximelagatran."
[Laughter.]
DR. SHETH: Can I now introduce Dr.
Jonathan Halperin, professor of medicine
at Mount
Sinai Medical Center in New York City,
and also the
Chair of the Executive Steering Committee for
SPF,
to give his views on the benefit/risk of
ximelagatran.
DR. HALPERIN: Thank you very much, Dr.
Sheth.
Mr. Chairman, members of the
committee,
ladies and gentlemen, I'm Jonathan Halperin
from
the Mount Sinai Medical Center in New
York City. I
direct a large anticoagulation clinic and
have a
longstanding research interest in
thromboembolism,
specifically in stroke prevention for
patients with
atrial fibrillation. I have served as a consultant
164
to AstraZeneca, as co-chairman of the
executive
committee governing the SPORTIF clinical
trials.
Now, Drs. Horrow and Sheth have
presented
the efficacy and safety data upon which
the sponsor
bases its application to use ximelagatran
as an
oral anticoagulant. My task is to put these data
into the perspective of clinical practice
and to
offer an assessment of the balance of
benefits and
risks associated with use of ximelagatran
for each
of these indications.
Let me begin by addressing the
hepatic
risk associated with ximelagatran. In these
trials, elevations in liver enzymes,
specifically
serum ALT above 3 times the upper limit
of normal,
typically occurred within the first 6
months of
treatment and were usually asymptomatic
and
reversible, even when the drug was
continued. In
my experience, these reactions are
manageable
because they arise early when
surveillance for
medication side effects is familiar to
most
physicians. Since the course was not invariably
benign, however, clinicians and the
manufacturer
165
must be vigilant to prevent more severe
liver
reactions, the frequency of which, like
many safety
issues, cannot be fully appreciated
before the drug
is brought to market and used more
widely.
A cautious algorithm and a risk
minimization program are necessary, and I
believe
they can be clinically achievable. With proper
use, the benefits of ximelagatran can
considerably
outweigh the risks.
The benefit/risk ratio is not a
fixed
property of any molecule. It varies depending upon
the patients in whom it's used, their
underlying
need for the therapy, and their
propensity for the
development of adverse events.
In the THRIVE III trial, for
example,
ximelagatran demonstrated convincing
antithrombotic
efficacy, achieving greater than 80
percent
relative risk reduction for the primary
endpoint of
symptomatic venous thromboembolism, with
bleeding
indistinguishable from placebo. In this
indication, as shown here, ximelagatran
demonstrates a positive net clinical
benefit.
166
From this and other recent
studies, the
advantage of extending anticoagulation
beyond 3 to
6 months to reduce recurrent thromboembolism
seems
clear.
My practice includes vascular medicine, and
I generally recommend continuation of
anticoagulant
therapy for patients with persistent
thrombotic
risk factors and for those with
idiopathic
thrombosis in whom no correctable cause
can be
identified and eliminated.
In chronic indications such as
this,
patients and physicians may find hepatic
surveillance with diminishing risk over
time
preferable to the coagulation monitoring,
the dose
adjustments, the dietary constraints that
are a
never-ending part of life with warfarin.
To complete the picture, I
await the
results of studies focused on the early
phase of
treatment of acute venous thrombosis with
ximelagatran.
Now, in the absence of
anticoagulant
prophylaxis, deep vein thrombosis
develops within
the first 10 days following knee
replacement
167
surgery in up to 70 percent of
patients. In the
United States, warfarin is the medication
most
frequently prescribed in this situation,
even
though its delayed onset leaves
anticoagulation
intensity deficient immediately after the
operation.
In the EXULT trials, the
efficacy of
ximelagatran was superior to
adjusted-dose warfarin
in patients undergoing orthopedic surgery
of this
type, and bleeding was comparable with
the two
strategies. Here, with 7 to 12 days of treatment,
the liver enzyme problem is not
seen. In this
indication as well, then, a net clinical
benefit
makes ximelagatran an appealing
alternative to
warfarin.
In the third indication,
prevention of
stroke and systemic embolism in patients
with
nonvalvular atrial fibrillation, previous
trials
have demonstrated that the high-risk
patients of
the sort enrolled in the SPORTIF trials
would face
an annual rate of stroke and systemic
embolism
around 8 percent without anticoagulation
and about
168
3 percent on warfarin. The SPORTIF results seem
convincing. Ximelagatran performed at least as
well as carefully modulated warfarin,
offering
consistent efficacy with a 1.6-percent
annual rate
of stroke and systemic embolism in both
trials.
There was no increase and perhaps even a
slight
decrease in bleeding complications.
Anticoagulation in the warfarin
arms was
about the best ever reported in a large
population
of patients with atrial
fibrillation. When
interpreting the results of the SPORTIF
trials,
therefore, it's important to bear in mind
that the
comparator was warfarin on its best
behavior.
Rates of intracerebral hemorrhage were
unusually
low compared with other anticoagulation trials.
From the perspective of a
treating
physician, an anticoagulant that reduces
the stroke
rate in high-risk cardiac patients at or
below 2
percent per year, with a low rate of
major
bleeding, while avoiding the complexities
of
warfarin, has until now been
unavailable. And when
tallying the risks and benefits of
ximelagatran,
169
remember as well that the outcomes we
seek most to
prevent--strokes and systemic embolic
events--are
devastating, often fatal.
Taking these together with
major bleeding
and mortality from any cause, including
hepatic
disease, there is to my mind a clear advantage
to
therapy with a direct thrombin
inhibitor. This net
benefit will weigh heavily with
clinicians caring
for patients with atrial fibrillation as
they
consider the trade-offs against warfarin.
One in four people over age 40
will
develop atrial fibrillation and face a
risk of
ischemic stroke about 5 times greater
than
individuals without this condition. Over half the
strokes occur in patients over 75 years
old, making
atrial fibrillation the most common cause
of stroke
in elderly women. As the population ages, the
number of patients with atrial
fibrillation at risk
of stroke is projected to double over
just two
generations.
Of the more than 2 million
Americans who
have atrial fibrillation today, about
half need
170
anticoagulation but fewer than a quarter
are
treated with warfarin. In primary care practice,
less than a quarter of patients on
warfarin have an
INR in the therapeutic range at any given
time.
Even in specialized anticoagulation
centers where
care is provided by carefully trained
nurses
following strict protocols and when
patients are
fully adherent, anticoagulation intensity
is seldom
maintained within the desired therapeutic
range
more than about half the time. No other drug has
generated such a burden just to support
its normal
use.
Based on all the available
evidence, I see
ximelagatran as a true therapeutic
advance, the
first oral direct thrombin inhibitor to
have been
extensively evaluated in clinical trials
using
fixed doses without coagulation
monitoring. In
these trials, ximelagatran proved an
effective
alternative to warfarin with a favorable
benefit/risk ratio even when the hepatic
findings
are taken into account. As more and more patients
need treatment, the quest for an
alternative to
171
warfarin has become urgent. On average, untreated
patients with atrial fibrillation face a
35-percent
lifetime risk of stroke, yet thousands
are
inadequately treated and hundreds of
strokes and
thromboembolic events occur across the
nation every
day that could be prevented with an
anticoagulant
safer and easier to administer than
warfarin.
The convenience of stable
dosing and
predictable anticoagulation with
ximelagatran will
appeal to patients and physicians alike,
potentially
enabling more to sustain anticoagulation
over time.
We stand now at the threshold
of a new era
in antithrombotic therapy. While other orally
effective alternatives to warfarin can be
expected
in the years ahead, none is even close to
coming
into hand for clinical use. There's a pressing
need for a better oral anticoagulant, and
ximelagatran has attracted the eager
interest of
all who share the goal of bringing
effective
therapy to many more patients to prevent
thousands
of thromboembolic events.
172
Thank you very much.
DR. BORER: Thank you very much, Jon.
You know, it may be, both in
the interest
of time and the interest of efficiency,
useful for
us to hear the first FDA presentation
right now
when Jonathan Halperin's summary is fresh
in
everyone's mind, because the first FDA
presentation
by Dr. Ruyi He is specifically about risk/benefit
assessment. Can we go ahead and do that then? And
we'll hold questions until we've heard
both.
DR. HE: Good morning--or I should say
good afternoon. I'm Dr. Ruyi He, medical team
leader in the Division of GI and
Coagulation Drug
Products, and I'll take the next 15
minutes to go
over my assessment of efficacy and safety
for
Exanta.
I will divide my presentation into two
parts:
short-term use and long-term use.
For each
part, I will discuss both efficacy and
safety
concerns.
Efficacy: short-term use in patients
undergoing total knee replacement
surgery. Exanta
was significantly better than warfarin
for primary
173
endpoint:
the incidence of total VTE and/or
all-cause mortality. However, this efficacy result
was driven by decrease in asymptomatic
distal DVT
which is not clinically meaningful. There was no
differences between groups for
symptomatic DVT,
proximal DVT, PE, or death.
This table was directly copied
from the
sponsor's submission. It shows no difference
between groups for death, PE, or
symptomatic DVT.
The main difference between groups was
the
incidence of symptomatic distal DVT, 19.2
percent
versus 26.7 percent in one study and 21
percent
versus 31 percent in the second study.
Now let's move to safety
assessment for
short-term use: bleeding events. A higher
incidence of bleeding events occurred in
the Exanta
group than in the warfarin group. This includes
both major bleeding and minor
bleeding. There were
two fatal bleeding evens, and both
occurred in the
Exanta 36-milligram group.
This table summarizes the
number of
patients with ALT elevation more than 3
times upper
174
limit of normal either during the 7 to 12
days of
therapy or 4 to 6 weeks follow-up
period. There
was no difference between groups during
the
treatment period; however, there was a
seven-fold
higher event rate in the Exanta group
than in the
warfarin group during the follow-up
period. There
was no long-term follow-up. The drug effect on
liver toxicity beyond 4 to 6 weeks is
unknown.
This table summarizes acute MI/CAD
events
that occurred in the two short-term
pivotal
studies.
Because the sponsor has shown this table
in her presentation, I will not go into
detail for
this table. The only thing I want to mention is
that the treatment group was well
balanced for
baseline condition, such as diabetes,
hypertension,
or coronary artery disease.
My safety concerns for
short-term use of
Exanta are: twofold higher incidence of major
bleeding events in the Exanta group; two,
higher
incidence of ALT more than 3 times upper
limit of
normal during the follow-up period, and
no
long-term follow-up data available;
three,
175
potential for duration of treatment to be
more than
12 days in clinical practice; and, four,
almost
threefold higher incidence of acute MI
event in the
Exanta group.
Now let's move to long-term
use, efficacy
in patients with atrial
fibrillation. The sponsor
conducted two noninferiority clinical
studies; one
was open label and the other was
double-blind. For
those two studies, the sponsor
prespecified
2-percent noninferiority margin was too
liberal,
and not formally agreed to by the
agency. The two
studies produced divergent results. Based on the
double-blind study, it could not be ruled
out that
the risk of stroke was twofold greater on
Exanta
compared to warfarin.
Long-term use, liver
toxicity. The
studies excluded patients who might have
potential
live impairment, such as patient who had
a liver
disease, ALT more than 2 to 3 times upper
limit of
normal, treatment with NSAID, or alcohol
abuse.
Before November 2001, liver function test
was as
follows:
monthly for 6 months, if ALT more than 3
176
times, then weekly; if more than 7 times,
Exanta
was stopped. After November 2001, because one
patient died from liver failure,
monitoring of
liver function test was enhanced to
monthly for 6
months; if ALT more than 2 times, then weekly;
if
more than 5 times or more than 3 times
for 4 to 8
weeks, Exanta was stopped. About 14 percent of the
patients being followed by later
monitoring.
This table summarizes the
number of
patients with ALT more than 3 times upper
limit of
normal alone or with both ALT more than 3
times and
a total bilirubin more than 2 times upper
limit of
normal.
This group of patients represent severe
liver injury. You can see that there was more than
a seven-fold higher incidence in the
Exanta group
than in the comparator group for ALT
elevation more
than 3 times, 7.8 percent versus 1.1
percent; same
as severe liver injury cases, 37 cases
versus 5
cases only. Nine out of 37 patients died, and
three of them may have been related to
Exanta. Let
me go over these three cases with you.
The first case was an
80-year-old man on
177
Exanta 36 milligrams twice daily for
atrial
fibrillation. At baseline, his liver function test
was normal; ALT was normal. Liver function test
was done monthly per protocol. The first month his
ALT was normal. The second month it was elevated
to 2 times upper limit of normal. However, the
third month his ALT was very high, up to
20 times
upper limit of normal. At this time Exanta was
stopped.
However, his condition did not better.
ALT continued to increase, total
bilirubin
increased, INR increased, and albumin
decreased.
This patient died from GI bleeding with
coagulopathy 2 months after Exanta was
stopped.
Autopsy showed a small, friable liver
with
extensive necrosis. This is the same as liver
biopsy report that was done a month
before his
death.
The most likely diagnosis for this case is
Exanta-induced hepatitis.
The second case was a
77-year-old man on
Exanta 36 milligrams BID for atrial
fibrillation.
At baseline, his ALT and albumin was
normal.
Albumin was 3.6. Liver function test was monitored
178
monthly per protocol. The first month his ALT was
normal; however, the second month ALT was
elevated
to 4.6 times upper limit of normal. Then per
protocol, liver function test monitoring
was
changed to weekly. Repeated result is unknown.
Two weeks later, the patient developed
bloody
stools.
At the emergency room, his blood pressure
was 76/45, and hemoglobin was 7. At this time PT,
PTT, and INR was all elevated. INR was 3.4.
The
plasma melagatran level was 0.25
micromolar, which
is in the therapeutic range. This level of
melagatran should not cause PT, PTT, and
INR
elevations. At the ER, his ALT was high, 569, and
the albumin was low, only 2.0, decreased
from
baseline 3.6 to 2.0. All of this indicates that
this patient had severe liver impairment
at this
moment.
The patient received transfusions.
After
transfusion at emergency, a gastroscopy
was done
which showed active bleeding site. The patient
received more transfusions; however, the
severe
coagulopathy had developed. This patient died a
few hours later. No autopsy was done, and in this
179
case was not included in the sponsor's
background
package.
The third case was a
73-year-old man on
Exanta 36 milligrams for DVT. His baseline ALT was
elevated, but less than 2 times upper
limit of
normal.
After a few days on Exanta, he fell sick
and the liver function test was done
early. At day
12, ALT was up to 4.5 times, and at day
18 up to
7.8 times. Hepatitis B was diagnosed on day 18,
and Exanta was stopped on day 24. After Exanta was
stopped, the patient's condition
continued to
worsen.
He developed hepatic encephalopathy and
died from liver failure 20 days later.
According to the literature,
mortality of
acute hepatitis B is between 0.1 percent
and 1
percent.
This table summarizes
discontinuation of
study drug due to an adverse event. A significantly higher
number of patients from the Exanta
group stopped Exanta due to an adverse
event; 4.6
percent in the Exanta group compared to
0.3 percent
in comparator group stopped study
medication
180
secondary to abnormal liver function
test.
Also, more patients in the Exanta
group
discontinued study treatment due to acute
MI and
bleeding event; 2.8 percent versus 1.9
percent for
acute MI; 1.2 percent versus 0.7 percent
for
bleeding event.
This table summarizes acute MI/CAD events
that occurred in the two proposed
long-term use
population, that is, patients with atrial
fibrillation or patients with VTE. The sponsor
also showed this table; however, I want
to make a
few comments on this table also.
First, I modified this table
and included
the population for post-acute coronary
syndrome.
Let's move to post-acute coronary
syndrome
population first.
The total CAD events in this
population
was 27 percent in the Exanta 36 milligram
BID plus
aspirin group. This is combination therapy. It's
a combination aspirin plus 36 milligrams
Exanta.
Compared to placebo plus aspirin, that is
30
percent in the placebo group. Two groups, no
181
significant difference. But this is combination
therapy.
And the second point I want to
make is the
baseline event rate for CAD was very high
for the
post-ACS population. The baseline condition,
baseline event rate was 27 to 30 percent,
and for
the atrial fibrillation population, also
baseline
event rates are high, that is, above 7
percent.
Because of this high baseline event, it
is
difficult to assess small proportion of
difference.
The third comment I want to
make is VTE
prevention study is a placebo-controlled
study.
You can see ten cases in the ximelagatran
group
versus only three cases in placebo group
for acute
MI event.
And, lastly, I want to point
out these two
populations is proposed in this three
indication,
two of long-term indication. This is the two
populations for the long-term use
population. In
the VTE population, it is significantly
higher for
both acute MI and CAD event.
Lastly, the treatment group
also was well
182
balanced for baseline condition that
included
diabetes, hypertension, and coronary
artery
disease.
My safety concerns for
long-term use of
Exanta are significant higher number of
patients
with severe liver injury after exposed to
Exanta,
include three deaths despite
protocol-specified
liver function test monitoring; higher
incidence of
withdrawal due to adverse event,
including acute MI
and bleeding events after exposed to
Exanta; and
higher incidence of acute MI in VTE
population,
including in placebo control study.
This concludes my
presentation. I want to
thank all the members of the Exanta
review team,
including especially those listed on this
slide.
DR. BORER: Okay.
Thank you very much.
It's 20 after 12:00, and we
must begin at
1 o'clock the open public hearing. So what I'm
going to do, unfortunately, is take a
break now for
lunch.
We'll have adequate time for questions
afterwards, and I'm sure there will be
many. And
we can complete the FDA presentation
after that,
183
and we'll go on to our open discussion.
So be back here by 1 o'clock at
the latest
because that's when we begin.
[Luncheon recess at 12:20 p.m.]
184
A F T E R N O O N S E S S I O N
[1:00 p.m.]
DR. BORER: We will reopen the meeting
with public comment, and before we begin
I want to
remind you that both the Food and Drug
Administration, that
is, the FDA, and the public believe in a
transparent process for information
gathering and
decisionmaking. To ensure such transparency at the
open public hearing session of the Advisory
Committee meeting, FDA believes that it
is
important to understand the context of an
individual's presentation. For this reason, FDA
encourages you, the open public hearing
speaker, at
the beginning of your written or oral
statement to
advise the committee of any financial
relationship
that you may have with the sponsor, its
product,
and, if known, its direct competitors.
For example, this financial
information
may include the sponsor's payment of your
travel,
lodging, or other expenses in connection
with your
attendance at the meeting. Likewise, FDA
encourages you at the beginning of your
statement
185
to advise the committee if you do not
have any such
financial relationships. If you choose not to
address this issue of financial
relationships at
the beginning of your statement, it will
not
preclude you from speaking.
With all that having been
said--and it's
quite a mouthful--why don't we begin with
the first
speaker.
DR. MESSMORE: Mr. Chairman and members of
the
Advisory Committee, as you can perceive, my
voice is not good today, but I will at
least
introduce my talk and I will be assisted
in this
endeavor by Dr. Wahi, who is a member of
our
research group at Loyola University Medical
Center.
I am Dr. Harry Messmore, M.D., FACP, and
I am
sponsored by the Thrombosis Hemostasis
Unit of
Loyola University Stritch School of
Medicine. This
unit has been in research for more than
30 years in
drugs of this nature and has been funded
with
various grants by low-molecular-weight
heparin
groups, pentasaccharide, and direct
thrombin
inhibitors. I am being paid for my expenses as
186
well as an honorarium since I am a
consultant to
that group, a senior member who is not
employed by
them at this time, but I am a volunteer
who is paid
a consultant fee periodically for special
studies
of this type.
I am a clinical hematologist
who has had
long experience in clinical and
laboratory
investigations of the
low-molecular-weight heparins
and antithrombin drugs, and I would like
to,
therefore, make a statement that we have
several--a
number of reservations regarding approval
of this
drug.
My formal statement then will be read by Dr.
Wahi, who will be able to, I think, get
the message
over more clearly than I can with my
voice today.
Thank you.
DR. WAHI: Mr. Chairman and members of the
committee, I am likewise sponsored by the
University of Loyola's thrombosis
research group.
We feel that there are some reservations
which we
need to bring to your attention.
We feel that for
the following reason the FDA should take
a good
look at this drug:
187
Number one, in patients who are
high risk
for orthopedic surgery for prophylaxis of
venous
thromboembolism in patients undergoing
total hip
replacement and total knee replacement,
ximelagatran in our opinion requires
further study
before it's approved for the following
reasons:
Number one, for both total hip
replacement
and total knee replacement, ximelagatran
should be
compared to low-molecular-weight heparins
for at
least 28 days for total hip replacement
and 17 days
for total knee replacement. The
low-molecular-weight heparins are
FDA-approved for
these indications and are in current
use. They
have been shown to be most effective when
used for
28 to 35 days in the case of total hip
replacement
and 17 days in the case of total knee
replacement.
Warfarin we do not think is a
good
comparator drug because it is inferior to
low-molecular-weight heparins in its
efficacy.
For secondary prevention of
venous
thromboembolism in patients after
standard
treatment of an episode of acute venous
188
thromboembolism, as studied in the THRIVE
III
trial, we feel that further studies are
required to
show that the abnormal liver function
tests, by
which we mean greater than 3 times normal
levels of
APLTs in 6.4 percent of the patient
populations,
are not--further studies are required to
show that
this elevation of enzymes are not
associated with
long-term sequelae.
T3B For preventing stroke and
other thromboembolic complications
associate with
atrial fibrillation, as studied in
SPORTIF III and
V trials, we feel further studies are
needed before
approval is considered. We base this opinion on
the fact that ximelagatran caused
abnormal liver
function in 6 percent of the patients in
the
reported study. The long-term consequences of this
should be studied before the drug is
approved in
this indication. Although it was judged to be
noninferior, these potentially harmful
side effects
will offset its noninferior status.
Furthermore, we feel that the
possibility
of hemorrhage due to increases in blood
levels over
189
time in a population whose renal function
is
deteriorating due to increased age or
other factors
would require periodic monitoring of its
anticoagulant effects, as well as the
monitoring of
renal function.
We also believe that evaluation
of drug
interactions with drugs such as
cholestyramine and
related drugs that could bind the drug
and prevent
absorption from the intestinal tract are
also
needed.
Furthermore, many of these
patients with
atrial fibrillation have associated
congestive
heart failure on the basis of coronary
artery
disease, which itself impairs hepatic
function and
may enhance the liver toxicity. We feel that
studies of the drug in this subset of
patients and
other patients who are at increased risk
for liver
disease, such as chronic hepatitis, et
cetera,
should be asked for.
In summary, we believe that
lack of
comparator studies with
low-molecular-weight
heparins for the high-risk period of 28
to 35 days
190
post-operatively in the total hip
replacement
patients and 17 days in the total knee
replacement
patients leaves the public uncertain as
to whether
ximelagatran is an effective and safe
prophylaxis
of venous thromboembolic disease in
high-risk
orthopedic surgery patients.
Furthermore, the second point
is that we
feel that the drug has not been adequately
studied
for long-term safety in atrial
fibrillation
patients, and its effectiveness as
compared with
warfarin is not sufficient to warrant its
use in
the presence of potential hazard in terms
of liver
damage.
Monitoring of liver function as
well as
the blood levels of the drug in the very
elderly,
over the age of 80, should be done before
it is
approved for this age group of Tunisia.
The last point I would like to
make is
that for the secondary prophylaxis of
venous
thromboembolic patients, in patients
previously
treated for an acute episode of venous
thromboembolic disease such as the THRIVE
III
191
trial, the issue of liver damage requires
consideration of whether there's clearly
sufficient
advantage over warfarin to warrant taking
the risk
of hepatic injury. A peer-reviewed publication of
the efficacy and safety of this drug,
given for the
first 6 months following an acute venous
thromboembolic episode is not yet
available to
permit me to comment on approving it for
this
indication.
It is also a concern that
physicians will
be using this drug off-label even if it
is approved
only for the patients already treated for
6 months
with warfarin.
Thank you very much for your
attention.
DR. BORER: Thank you very much, Dr. Wahi
and Dr. Messmore.
Before we go to the next
speaker, I have
one announcement. For recording purposes, please
speak into the microphone. Anyone who speaks, get
the microphone close to you, if you need
to. And
please silence the cell phones, if
anybody has one
in here.
It's heard on the recording and makes it
192
hard for the transcriptionist to
transcribe the
meeting.
Okay. Can we have the next, speaker
number two, please?
MR. BARANSKI: Good afternoon, Chairman
and committee members. My name is Jim Baranski. I
am the CEO and executive director of the
National
Stroke Association. I would like to point out that
I am here, as they say, on my own
nickel. However,
the National Stroke Association, like
most
not-for-profit associations, derives its
funding
from community-based projects from the
government,
as well as from industry, of which
AstraZeneca is
one of our many industry sponsors.
For those of you who may not be
familiar
with the National Stroke Association, we
began our
mission in 1984, and that mission was to
reduce the
incidence and impact of stroke. Over the years, we
have developed, with the help of many of
the
leading thought leaders of stroke,
programs to
better educate the professional community
as well
as the public.
193
In 1999, we were the first ever
to produce
primary prevention guidelines. Those guidelines
were published in JAMA. At the
end of this month,
we will be publishing secondary recurrent
guidelines. Those will be published in the Journal
of Stroke and Cerebrovascular Disease.
I guess, you know, after hearing
a number
of the comments earlier, please
understand, I am
not here to debate the data, the
difference between
1.6 or 1.4 percent. Really, I am here to express
the concern that the patient
constituents, as well
as the professionals that we represent,
have a
tremendous concern over the current
problems of
warfarin.
As we all know, the issues of monitoring
INR, I can tell you that when I speak to
patients,
the cattle rancher in the middle of Montana,
monitoring INR for him is a great
challenge. And
you know what? I'm not so certain that his
professional is that confident about
their ability
to monitor INR. It seems to me that the most
valuable aspect of an opportunity here
today is
this whole ability to try to close the
gap that
194
currently exists. That gaps that exists is really
50 percent of AF patients are being treated
currently on anticoagulants. Only 50 percent, out
of the total population of AF that I
don't think
we've mentioned today, there's 2.2
million
Americans currently that struggle with
this
problem.
So to our way of thinking, if
we have this
50-percent gap that exists, and if one of
the
reasons--one of the reasons that we're
hearing from
our constituency is because of the whole
problem of
monitoring, of INR monitoring, and the
fear that
has developed over the past 60 years,
then we
embrace this as an opportunity or any
other
therapeutic opportunity to try to bridge
that gap,
to remove the fear, to try to help create
better
patient treatment for a population that,
by and
large, is aging quickly and it's very
difficult for
these people to maintain the patient
medications
that are currently available.
Thank you.
DR. BORER: Thank you very much, Mr.
195
Baranski.
Let's go on to the third
speaker.
DR. ANSELL: Mr. Chairman and members of
the committee, good afternoon. My name is Jack
Ansell, and I am a practicing
hematologist and a
professor of medicine at Boston
University School
of Medicine, where I also direct the
anticoagulation service
at Boston University Medical Center.
I've spent the last 25 years or so
treating
patients with thromboembolic disease and
managing
patients on oral anticoagulants. I'm also the
founder and Chair of a professional
group, the
Anticoagulation Forum, which is a network
of over
3,000 health care providers who manage
oral
anticoagulation in anticoagulation
clinics and
interface with somewhere between 400,000
and
500,000 patients on oral anticoagulants.
In the spirit of full
disclosure, let me
first say that I consult for AstraZeneca and
have
research support from AstraZeneca, the
company
whose product is being discussed
today. But I also
consult and receive research support from
a number
196
of other competing pharmaceutical
companies in this
field.
For me it is not the company but the
products that are being developed that is
important.
I'll also say that I am here on
my own
account to testify and to make this
statement
today, with no honorarium or expense
coverage from
AstraZeneca.
I am not here today to comment
simply on
the attributes of a new
anticoagulant. I also want
to emphasize just how problematic
existing therapy
is with oral anticoagulants. And, finally, I want
to comment on how important today's
review will be
in the context of a number of new
anticoagulants
that will come before this group sometime
in the
next 5 to 10 years.
One not need be in the medical
profession
to know something about Coumadin, the
oldest and
the only oral anticoagulant
available. There is
probably no more than one degree of
separation
between anyone on this planet and someone
who
receives Coumadin therapy or a similar
agent. And
197
for those more intimately familiar with
this drug,
they know the vicissitudes involved in
taking
Coumadin.
It is not an easy drug to take and
certainly not an easy drug to
manage. Warfarin,
the generic name for Coumadin, is a drug
that under
the most ideal conditions would probably
never come
close to approval if it were under review
today.
What, then, is wrong with
warfarin as a
therapeutic modality? Let me summarize. Warfarin
has many drawbacks. You've heard a number of them
today.
It has an unpredictable response in
therapeutic levels that requires the need
for
frequent venipunctures, frequent INR
monitoring and
dose changes, frequent visits or contacts
to a
health care provider to manage that
therapy.
Warfarin interacts with multiple drugs
and foods
that increase the risk of bleeding or
thrombosis
and affects the quality of life.
Warfarin has a narrow
therapeutic range
such that the bleeding risk increases
significantly
with too high an INR, and the risk of
thrombosis
increases when the INR is too low.
198
Warfarin is associated with a
very high
incidence of major bleeding or thrombosis
that is
not reflected by the clinical trials that
many of
us are familiar with. In the real world of
warfarin management, rates of major
bleeding and
thrombosis combined are as high as 15
percent per
year of treatment. Imagine going to your physician
with new onset atrial fibrillation and
having your
doctor say, "Here, take this pill to
prevent a
stroke.
But, by the way, there may be something
like an 8- to 15-percent chance of you
having a
major complication or life-threatening
complication
with this drug."
Warfarin therapy is associated
with an
impaired quality of life for both the
patient and
the
doctor as well, and it's one of the most common
causes of malpractice litigation.
In order to optimize
anticoagulation
management and to prevent complications,
an entire
industry has sprung up around oral
anticoagulation.
This is the industry of anticoagulation
clinics,
and it is my organization that exists to
serve
199
those providers who work in these
clinics.
What do we need to improve
anticoagulation
therapy?
Well, we need new drugs that work in
different ways. We need drugs where the
therapeutic response is predictable and
monitoring
is not required. We need new drugs that have few,
if any, interactions with other drugs or
with
foods.
We need new anticoagulants that have a
relatively wide therapeutic window so
that the risk
of bleeding or thrombosis is
minimized. And, most
importantly, we need new agents that are
available
in oral formulation so that we can reduce
hospitalizations and the need for daily
injections,
whether in the hospital or out of the
hospital.
So I am not here to simply
speak for the
drug under review today. Ximelagatran has
performed exceedingly well in extensive
clinical
trials, and as a physician, I have full
faith in
using Exanta for the indications being
discussed.
But ximelagatran is simply the first of a
number of
new oral anticoagulants that will come
before the
FDA in the next several years, and these
drugs must
200
be viewed in the context of what they are
intended
to replace.
One must balance the needs for
a better
therapy, even if that therapy is not
perfect,
against the very imperfect and often
dangerous
therapy we currently use, and in some
cases don't
use because of fear of complications.
For me, it's an exciting time
to be
working in this discipline. We are on the cusp of
a revolution in anticoagulant
therapy. Ximelagatran
represents the first new oral
anticoagulant
since the discovery of dicumarol in
1940. Exanta
is poised to have a major beneficial
impact on the
outcomes of patients currently taking
Coumadin, on
their quality of life, and on the overall
cost of
health care.
Thank you very much.
DR. BORER: Thank you very much, Dr.
Ansell.
Can we have the fourth speaker,
please?
DR. LURIE: Good afternoon. I'm Peter
Lurie.
I'm a physician with Public Citizens Health
201
Research Group. We take no money from industry,
from government, or from any professional
organizations.
Our remarks this morning will
take the
form of a brief summary of the safety and
efficacy
data that exist for each of the three
indications,
and then afterwards, I'll talk briefly
about the
risk management program that has been
proposed by
the sponsor.
In summary, our position is
that for the
first and third of these indications,
there are, in
fact, no convincing evidence--no
convincing data
showing the effectiveness of the drug for
those two
indications, and we're unconvinced that
for the
second, the long-term prevention of
venous
thromboembolus after standard treatment,
that the
risk management program so far proposed
is going to
be adequate to reduce the risks
particularly to the
liver associated with that indication.
Let's go through the first one,
by which I
mean the knee replacement
indication. The
sponsor's data indicate that the drug
failed to
202
reduce the incidence of symptomatic
proximal or
distal deep venous thrombosis compared to
warfarin.
There is an impressive-appearing
reduction in the
incidence of asymptomatic DVTs, but
closer
inspection reveals that essentially all of that
reduction occurred in asymptomatic distal
DVTs that
were diagnosed only by venography. As the medical
officer concludes, this "is not
clinically
meaningful."
In addition, although there are two drugs
that are approved for this indication,
the sponsor
chose instead to compare ximelagatran to
warfarin,
which is not approved in this country for
this
indication. This seems like an unfair comparison.
Warfarin takes 3 to 5 days to even reach
therapeutic levels compared to merely
hours for
ximelagatran. And, indeed, the medical officer
said this comparison is unfair,
particularly
because the study only lasts 7 to 12
days, and so
many of the warfarin patients were not
adequately
anticoagulated.
Moreover, any convenience
advantage over
203
approved medications conferred by
ximelagatran
being an oral medication is really
diminished in
this short-term, substantially inpatient
setting.
But even in the short-term
studies, there
were some safety concerns. Major bleeding was
increased in the ximelagatran-treated
group,
although not statistically significantly
so. But
there was a statistically significant
increase in
coronary artery disease events which were
described
by the medical officer as
"unexpected and
worrisome." In addition, the rate of ALT was
increased both at the end of treatment
and, more
ominously, 4 to 6 weeks after the
treatment had
been completed.
In sum, ximelagatran's efficacy
appears to
be limited to asymptomatic distal DVTs,
and even
then, only after comparison to a drug
unapproved
for this condition. It's a classic technique if
you're going to do an active controlled
trial to
pick the comparator that is likely to
benefit your
active drug, and I think that's what was
done here.
Given concerns about the
propensity of the
204
drug to induce bleeding, cardiovascular
events, and
ALT abnormalities, and the existence of
other
approved medications for this condition,
we do not
believe that ximelagatran should be
approved for
this condition.
As to the second indication,
long-term
secondary prevention of VTE after
standard
treatment, there's only a single 18-month
placebo-controlled trial, THRIVE III,
which has a
decidedly modest 73-percent follow-up
rate.
However, although it has no impact upon
mortality,
ximelagatran did show an impressive
reduction in
both symptomatic VTEs as well as
pulmonary embolus.
And so we do think that it appears to be
superior
to placebo for this condition.
However, one then needs to
think about the
safety issues, and I'm going to discuss
the safety
issues for this and the third indication
together.
For all the patients receiving
ximelagatran on a long-term basis, 7.6
percent
developed ALTs greater than 3 times the
upper limit
of normal, 7 times higher than the case
in the
205
comparators group. It is noteworthy that in
trials--let's step back and say let's
compare this
to the situation with troglitazone or
Rezulin.
Those were typically 6-month studies,
which is in a
way analogous to the situation here
because the
great majority of hepatotoxicity does occur
within
the first 6 months of treatment.
Troglitazone caused an increase
above 3
times the upper limit of normal in only
1.9 percent
of patients compared to 0.6 percent on
placebo.
Compare here to 7.6 percent with 1.1
percent on the
comparators. That drug, of course, eventually got
banned, but only after there had been 94
cases of
liver failure, most of them fatal.
More serious hepatotoxicity,
defined as 3
times the upper limit of normal or twice
the upper
limit of normal for bilirubin, were
observed in 37
patients, which works out to a relative
risk of
6.6-fold over the comparators. And as you know,
nine patients died and three were judged
by the
medical officer to be related to
ximelagatran,
which works out to about 1 in 2,000
patients. This
206
rate will likely be higher in clinical
practice as
patients excluded from the clinical trials
come to
be treated with ximelagatran. And not so far
emphasized in this hearing but available
in the
medical officer's summary is additional
information
about how high those elevations of ALT
can be: 4.3
percent had ALTs greater than 5 times the
upper
limit of normal, 0.4 for comparators, and
1.6
percent had ALTs greater than 10 times
the upper
limit of normal, 0.1 percent for
comparators. And
as we have seen, there is no subgroup
that
adequately can be predictive of where the
toxicity
will occur.
For the second indication,
there is a
non-statistically significant increase in
coronary
artery disease for both long-term
indications. So,
in sum, for the second indication, while
the drug
appears to be effective, although not
reducing
mortality, the risks are significant and
I'll talk
about the risk management policy in a
moment.
Finally, the third indication
related to
atrial fibrillation. Two studies have been
207
submitted, and in our estimation both are
seriously
flawed.
A noninferiority trial is in conceptual
terms, I think, the right way to go for
this
indication, because I don't think it
would be
ethical to randomize people to
placebo. But if you
do it, you have to do it right. And in this case,
the company has prespecified a
noninferiority
margin of 2 percent. The medical officer says that
this 2-percent margin was "not
agreed to by the
agency," and, moreover, "the
margin chosen was too
liberal," rather different than what
the company
told us this morning.
The two studies, despite being
similarly
designed, had divergent results, as was
pointed out
this morning, and the relative risks,
which I
calculated to be the same as you, were
0.71, i.e.,
that ximelagatran looked better in the
less
important open-label study, and 1.39,
i.e.,
ximelagatran looked worse in the more
important
double-blinded study. The statistical reviewer
correctly states, "In general, the
results from the
double-blind studies are more reliable
for many
208
reasons." And then he goes on to say, "Unless the
clinical judgment is that a loss of 2
percent"--the
noninferiority margin--"of the
effect of warfarin
is clinically acceptable"--and I
would just add, 2
percent in the context of very varying
rates with
respect to warfarin alone shown on one of
those
slides.
"unless the clinical judgment is that a
loss of 2 percent of the effect of
warfarin is
clinically acceptable, in my opinion
ximelagatran
has not been demonstrated to be
noninferior to
warfarin." I think that's absolutely right.
So, in view of the significant
problems
with trial design and the toxicity which
I've
already described with respect to
long-term use, we
don't think it should be approved for
this either.
The sponsor has proposed an ALT
monitoring
program that is similar to that adopted
in the
clinical trials, but despite that, cases
of severe
liver injury, including fatal ones,
occurred even
with levels of compliance that are likely
to have
been much higher in the clinical trials
than they
will be in actual practice. And if we look again
209
at the situation with Rezulin, which had
a liver
function test requirement, but the third
month of
therapy fewer than 5 percent of patients
were
getting the required LFT monitoring. So that's of
great concern as well.
FDA's Office of Drug Safety
concludes, "We
do not agree that the sponsor's proposed
Risk MAP
program is adequate. Currently, the proposed
monitoring plan provides no guarantee of
safeguarding the patient from developing
a rapid
onset and life-threatening reaction."
It's also true that the risk
management
program does not even begin to address
the problem
of delayed hepatotoxicity in the
prophylactic use
after surgery indication. And, furthermore, it
makes no attempt to reduce the risk for
bleeding or
coronary artery disease.
There are a number of
approaches that go
beyond the Risk MAP or LFT monitoring
approach that
have been put forth for consideration by
FDA, and
if the drug is approved--which, as I've
said, for
two out of three indications, at least,
it should
210
not--these should be implemented, and
mention has
been made of a black box warning,
mandatory patient
registries for long-term users that would
be linked
to performance, physician-patient
agreements, and
restrictions on promotion, distribution,
and
packaging.
Thank you very much.
DR. BORER: Thank you, Dr. Lurie.
Can we go on to the fifth
speaker, please?
MR. LODWICK: I'm Al Lodwick. I'm a
pharmacist, certified anticoagulation
care provider
in Pueblo, Colorado. I have been a consultant for
AstraZeneca, but currently have no
contract with
them, nor any other financial interest in
this
product.
You might argue that I have some
financial interest in that I spend my
full-time
monitoring warfarin, but that's beyond
the scope of
this.
I also own a website,
warfarinfo.com, that
puts me in contact with people around the
world
with anticoagulation and its related
problems.
A replacement for warfarin has
long been
211
awaited.
Exanta has generated a great deal of
anticipatory excitement. The fact that it will not
have to be monitored had been repeated so
often
that it has become almost a mantra.
In the published premarketing
studies,
approximately 6 percent of the
participants
experienced some, at least transient
elevations of
liver function tests. Please notice I said
"published premarketing
studies." The Wall Street
Journal of September 8, 2004, published
an article
about Forest Laboratories settling a
lawsuit
brought by the New York Attorney General
over
suppressed data concerning several of
their drugs.
The same article mentions GlaxoSmithKline
settling
a lawsuit brought by the same party for
concealment
of data about Paxil. Meanwhile, Eli Lilly and
Merck & Company have announced plans
for a clinical
trial databases. This raises the issue of whether
or not there are unpublished studies
about Exanta
that could be pertinent to this Advisory
Committee's decision, and I suggest that
the
committee ask that question rather than
wait for an
212
Attorney General to ask it.
While I have no insight as to
what this
body will recommend or what the final
decision on
Exanta will be, I'm assuming that it will
eventually be approved for marketing in
the United
States with some requirement for
monitoring liver
function tests. History provides us with some
rather troubling incidents where recommendations
of
the FDA have been widely ignored by
clinicians. In
view of this, in the near religious
fervor that
this drug will not require monitoring, I
request
that the Cardiovascular and Renal Drugs
Advisory
Committee recommend that any liver
function
monitoring requirement be stated in the
strongest
possible terms in the labeling of
Exanta. In
addition, all advertising material should
boldly
state this requirement.
Consideration should also be
given to a
warning label for patients stated,
"You should have
had a liver function test before you
start taking
this medication, and this should be
repeated every
month for the next x months. Failure to do so
213
could result in potential liver problems
being
overlooked."
Thank you for your time and
consideration.
DR. BORER: Thank you, Dr. Lodwick.
We'll go on to the sixth and
final
speaker?
DR. COLGAN: Mr. Chairman, committee
members, and members of the FDA, my name
is Kevin
Colgan, and testifying with me is Victor
Tapson,
who is sitting back in the front
row. Dr. Tapson
is professor of medicine at Duke
University School
of Medicine and practices in the Division
of
Pulmonary and Critical Care Medicine at
Duke
University Medical Center. I am Vice President for
Outcomes Research at EPI-Q,
Incorporated. I am a
pharmacist with other 20 years of
practice
experience in hospitals and health
systems. And I
am a member of the board of directors of
the
American Society of Health-System
Pharmacists.
However, we are here today
representing
our fellow researchers from the National
Anticoagulation Benchmark and Outcomes
Report
214
Steering Committee, also known as the
NABOR
Steering Committee. They include Richard Becker,
who is a cardiologist also from Duke
University
Medical Center; Dr. Albert Waldo, who is
an
electrophysiologist from Case Western
Reserve
University and the University Hospitals
of
Cleveland; Dr. Joseph Caprini, who is a
vascular
surgeon from Evanston Northwestern
Hospital and the
Feinberg School of Medicine; Dr. Thomas
Hyers, who
is an internist from St. Louis University
School of
Medicine and CARE Clinical Research; Dr.
Richard
Friedman, who is an orthopedic surgeon
from the
Charleston Orthopedic Associates and the
Medical
University of South Carolina; and Dr. Ann
Wittkowsky from the University of
Washington School
of Pharmacy and Medical Center, director
of the
anticoagulation clinic; Dr. Agnes Lee,
who is a
hematologist from McMaster University;
and Dr.
David Ballard and Dr. Roger Khetan. Dr. Ballard is
head of quality improvement at Baylor
Health Care
System, and Dr. Khetan is a hospitalist.
Over the past 18 months, the
NABOR
215
Steering Committee has performed an
anticoagulation
benchmarking study in U.S.
hospitals. We have
studied atrial fibrillation, acute
myocardial
infarction, deep vein
thrombosis/pulmonary
embolism, and venous thromboembolism
prophylaxis
for total knee replacement, total hip
replacement,
and hip fracture repair surgery. Our study was
funded with an educational grant provided
by
AstraZeneca. They had no role in the design or the
conduct of the study, collection,
management,
analyses, or interpretation of the
data. In
addition, we have funded the cost of
attending this
open hearing ourselves--we are here on
our own
nickel--because we are genuinely
overwhelmed by the
degree of inadequate treatment we
observed in our
study.
Therefore, we sincerely appreciate the
opportunity to comment at the Advisory
Committee
Open Hearing for ximelagatran.
The NABOR Steering Committee
brings two
perspectives to the issues being addressed
today.
First, we bring the perspective of
researchers. We
have each been involved in
anticoagulation and/or
216
process improvement research and espouse
the
practice of evidence-based antithrombotic
therapy
through the implementation of
professional
guidelines published by the American
Heart
Association, the American College of
Cardiology,
the American College of Chest Physicians,
the
American Society of Health-System
Pharmacists, the
European Society of Cardiology, and many
other
professional groups.
Second, we are clinicians who
daily care
for patients requiring anticoagulation
therapy. We
realize the necessity of anticoagulation
therapy
and experience the real world benefits
and the
shortcomings of warfarin, which has been
our
primary oral weapon against thromboembolic
disease
for decades.
We understand that the Advisory
Committee's principal concerns are
efficacy and
safety.
However, there is a third dimension that
we would like to discuss today, and that
is the
dimension of both the non-treatment and
sub-optimal
treatment of patients at risk of stroke
and venous
217
thromboembolism. And to show you the magnitude of
this problem, presently there are 4.7
million
stroke survivors living in the United
States with
15 to 30 percent of those being
permanently
disabled.
Stroke costs the United States $31
billion in direct costs and $20.2 billion
in
indirect costs annually. Fatality from thrombosis
is about four times as prevalent as
fatality from
malignancy.
Our NABOR study was performed
at 38 U.S.
hospitals from 28 different states and
included
3,778 randomly selected record for
patients
primarily treated during the calendar
year 2002.
It was a retrospective, benchmarking
study that
included 21 teaching hospitals, 13
community
hospitals, and four veteran
administration
hospitals.
Of the 945 patients with a
primary or
secondary diagnosis of atrial
fibrillation, 86
percent were stratified as having a high
risk of
stroke using the American College of
Chest
Physicians risk stratification
scheme. Of those
218
stratified to high risk, only 55 percent
received
warfarin, which has been shown to reduce
stroke by
approximately 65 percent in unselected
patients
with atrial fibrillation. Interestingly, this
result confirms other studies dating as
far back as
1980 and is practically the same at that
published
by Jencks, the Centers for Medicare and
Medicaid
Services in 2003 for results compiled
from 1998 to
2001.
Furthermore, neither warfarin nor aspirin
were prescribed in 21 percent of
high-risk
patients, including 18 percent of those
with
previous stroke, transient ischemic
attack, or
systemic embolic event, despite 43
percent having
no identifiable bleeding risk. We felt it logical
to expect that atrial fibrillation
patients with
previous stroke, TIA, or systemic embolic
event,
and thus the highest annual stroke rate,
would be
much more likely to be treated with
warfarin.
However, only 61.2 percent received
warfarin. This
was disturbing, considering that the
number needed
to treat to prevent just one stroke in
this
population ranges from just 7 to 10.
219
Of the 939 patients with deep
vein
thrombosis and/or pulmonary embolism, we
found
inadequate treatment overlap of a heparin
compound
and warfarin, which included only one in
four
patients who had at least 5 days of
concomitant
therapy.
We found a tendency to delay discharge in
isolated DVT patients rather than
providing
ambulatory bridge therapy which resulted
in a
3.8-day lower length of hospitalizations
in those
who received it. And we found inadequate or lack
of treatment at discharge. Ten percent of DVT
patients were discharged without a
prescription for
an oral or an injectable anticoagulant
after
receiving a mean duration of only 10.6
days of
treatment during hospitalization, many of
whom were
diagnosed with idiopathic disease or who
had
previous DVT or malignancy. In many of these
patients, we would expect an extended
duration of
anticoagulation therapy for at least 6
months or
longer.
Of the 928 patients who had
undergone
orthopedic surgery, 38 percent underwent
total knee
220
replacement, 30.6 percent underwent total
hip
replacement, and 31.4 percent underwent
hip
fracture repair. Surprisingly, 14.4 percent
received inadequate prophylaxis with
aspirin or no
prophylaxis at all; 6.1 percent received
nothing.
Total knee replacement patients received
a mean of
3.2 days of post procedure
anticoagulation in the
hospital; however, 17 percent did not
receive a
discharge prescription for
anticoagulation.
Likewise, 17 percent of total hip
replacement and
47 percent of hip fracture repair
patients did not
receive a discharge prescription for
anticoagulation after a mean of only 3.4
and 4.7
days of post procedure anticoagulation,
respectively.
Now, I've given you a lot of
data, but let
me interject that the NABOR Steering
Committee
members are very concerned about the
magnitude of
under-treatment we observed in our
study. And if
it is representative of treatment in the
United
States, it represents a substantial
public health
problem.
Despite its enormous preventive potential
221
oral anticoagulation is prescribed for a
little
more than half of high-risk atrial
fibrillation
patients.
At least one in ten DVT and pulmonary
embolism patients and one in five
orthopedic
surgery patients do not receive adequate
anticoagulation and/or duration of
therapy. This
level of non-treatment and sub-optimal
treatment in
part shows the real-world limitation of
anticoagulation with warfarin. After decades of
research on the benefits of warfarin to
prevent
stroke in atrial fibrillation and to
treat and
prevent venous thromboembolism,
clinicians still
don't use the drug at all or use it
incorrectly.
We feel that having an alternative oral
anticoagulant could increase the number
of eligible
patients who are treated and hopefully
reduce the
death and disability caused by
cardioembolic stroke
and venous thromboembolic events.
Lastly, anticoagulation always
has
involved a balance of risk and
benefit. The
SPORTIF, THRIVE, and EXULT randomized
controlled
clinical trials have demonstrated the
efficacy of
222
ximelagatran for preventing blood
clotting.
Ximelagatran does not have to be
constantly
adjusted, may have less risk of bleeding
and a
lower potential for drug interactions
than
warfarin; however, it requires monitoring
of liver
function.
We believe the balancing of risk and
benefit would be enhanced if another
effective oral
anticoagulant with a tolerable risk
profile was
available for prescription in the United
States.
Having two oral agents, namely, warfarin
and
ximelagatran, would provide practitioners
greater
flexibility in best matching an oral
anticoagulant
with the unique needs of every patient.
On behalf of our steering
committee, we
thank you for the opportunity to comment on
our
concerns for primary and secondary
prevention of
thromboembolic disease and how our
research
findings weigh in the consideration of
the New Drug
Application for ximelagatran.
Thank you all very much.
DR. BORER: Thank you very much, Dr.
Colgan.
223
Is there anyone else who hasn't
registered
who has a statement during this open
portion?
[No response.]
DR. BORER: Okay.
If not, we'll move back
to our agenda. We're going to follow this plan:
First, I short-circuited a question by
Dr. Sjogren
earlier, and if you want to raise that
question
now, we'll start with that. Then we'll go on to
allow Dr. Sheth to answer the questions
that we
raised earlier that I think she has
answers to now.
Then we can have questions regarding the
risk/benefit analysis from the entire
committee for
Dr. Halperin and for Dr. He, and then
we'll
complete the FDA presentation with any
questions
that we have that follow, then the charge
to the
committee, and then we'll close the
meeting to
non-committee speakers and go through our
analysis
based on the questions we've been given,
which
should bring us right up to about one
minute to
5:00.
Dr. Sjogren, do you want to go
ahead?
DR. SJOGREN: My comments have to do with
224
the liver test, particularly ALT, and the
FDA
description of the three patients that
died I think
answers my question in a better way than
earlier,
because that's exactly what I was
asking: What
happened to the ALTs over time? And I calculated
that some patients that started with
normal ALTs
promptly went into major abnormalities in
12 days,
in 30 days, in a very short time. And so the
proposed monitoring for these patients
would not
touch--would not have caught these
patients
because, you know, they expect to be an
ALT that is
twice the normal and then go on with
monitoring.
So I'm still very concerned
about that,
and I would think that as a liver doctor,
if I saw
any abnormality, I would discontinue the
drug
immediately based on what the FDA has
shown us.
And
I would prefer that patients have an ALT to
begin with before they start the drug,
and if it is
abnormal, then I would not feel safe to
give the
drug.
I think there are some patients that just
the toxicity progresses very fast and
we're unable
to tell who those are.
225
DR. BORER: I should point out we don't
actually know what the algorithm is that
the
company ultimately will come up
with. Their Risk
MAP is something that they'd have to talk
to the
FDA about. But all these points that you've made I
think we have to keep in mind.
Okay. Dr. Sheth, as you get prepared to
respond to the several questions, I found
in my
briefing document the areas that had
caused me to
raise the question about rebound, and
perhaps so
that everybody can look at these data,
they're on
page 47, Table 18 of the FDA briefing
document.
The FDA briefing document. Oh, I'm sorry. You're
quite right, there are. This is called "Clinical
Review DGICDP." And the next is "Clinical Review
DCRDP" and that is Table 45 on page
75 of that tab.
And the only reason I note these is that
it was
from these tables that I made the
calculations that
led to the question. You may want to look at the
primary data that generated the question.
Okay. Dr. Sheth?
DR. SHETH: It's not a very good
226
reproduction. Is this the table that you're
talking about?
DR. BORER: Yes, indeed.
DR. SHETH: Okay.
All right. Let me do
this.
The data that we prepared looks at the
number of myocardial infarctions that
occurred
after treatment over those that occurred
during and
after treatment. I think that was one of the
questions prior to the break. And if I could start
first with the long-term exposure pool,
and then
I'm going to move actually to the
warfarin
comparison pool, which is, I think, the
real
question of interest, so that would be
O-49,
please.
Okay.
Just focus on this bottom line,
total MIs,
and this should be the number of
myocardial
infarctions that occurred after study
treatment
stopping over the total number in the
program, and
that would be--for the long-term pool,
it's 28
percent versus 23.
Can I have O-48, and let's do
the surgical
population, please?
227
T4A Here, on this bottom line, the
patients with an MI after stopping
treatment, it's
broken up into 36 and 24 milligrams, over
the total
during the program, and it's four out of
ten for 36
milligrams, one out of seven for 24 milligrams,
and
two out of four for warfarin, and one out
of two.
The percentages are listed here, but I
just caution
that the numbers are very small in this
analysis.
I'd also like to ask Dr. Peter
Kowey to
come and discuss this a little further,
if I may.
DR. FLEMING: While he's doing that, can
you leave the previous slide up?
DR. SHETH: O-47?
I think the one just
before, is that correct?
DR. KOWEY: This problem with myocardial
infarction and coronary events that
occurred in the
orthopedic surgery population was a
particular
concern to us when we learned that there
was a
concern by the agency. Obviously, those data
weren't quite known to the sponsor and to
the
consultants. And the reason why it was a surprise
to us is because, under ordinary
circumstances,
228
when we're confronted with a question
having to do
with something like an acute ischemic
syndrome,
acute ischemic events, and you ask us the
question
could this drug somehow have been causing
an issue,
the reaction is go study the highest-risk
patient
population and show us the data in a
population of
individuals who are at maximum risk. That's an
acute coronary syndrome population.
And so we thought and I still
think that
the data from ESTEEM really provides the
most
information that we have about the
question of
whether this drug under some
circumstances could be
causing an acute coronary syndrome. Keeping in
mind the fact that ESTEEM finished with a
14-day
wash-out period during which patients
were still
counted in the trial and events were
still counted
in the trial. And even with that, there was no
signal that there was an excess risk of
having
acute coronary events, and obviously
there was an
advantage to ximelagatran.
So the answer to the question
of what does
it mean when you see a short orthopedic
experience
229
study where you see this excess event
rate.
Clearly, we agree with you that that's a
different
setting, and the patients are freshly
post-operative, and there may be
something uniquely
susceptible about those individuals. But with
those kinds of numbers and the very small
numbers
of events that were seen in those trials,
in the
setting of having a very definitive
clinical
trial--and, again, that's the reaction
that I would
have, is go do a study in the maximum
at-risk
population. You have those data. It seems to
me--in addition, in that study you have
withdrawal
data.
It seems to me that that probably answers
the question.
Therefore, I think that what
you're seeing
in the orthopedic surgery experience is
not real
and that it represents a play of chance.
DR. BORER: Okay.
I'll do something that
I shouldn't do, which is to tell you my
bias, which
is that I agree with you about the
importance of
ESTEEM.
But the question was specifically not
about total MI risk. It was about the potential
230
for rebound that might have to be
remediated with
some adjunctive therapy if this drug were
approved
and used.
DR. KOWEY: Yes, I agree with you, and I
think there are two pieces of information
that gave
me reassurance on that because, Jeff,
honest, we
thought the same thing. One piece of information
is that in SPORTIF there were patients
who went on
and off the drug on a regular basis. As you know,
this drug is not indicated for
cardioversion, for
example, so there were patients that were
going on
and off the drug constantly in
SPORTIFs. And in
SPORTIF, as you say, in an
adjudicated--that's the
other thing, by the way. I was a little bit
surprised that we weren't giving a whole
lot more
weight to adjudicated events than we were
to
investigator-reported events, which we
traditionally don't really do. I pay a lot more
attention to what an adjudication panel
blinded
would do with this.
The other piece of information,
as I said
earlier, is that also in the ESTEEM
experience,
231
there was a withdrawal phase so that we
do know
that we have an opportunity to really
observe the
maximum risk population coming off the
drug without
an excess event rate.
Those two pieces of information
for me
were very reassuring and, again, that's
why we were
a little bit taken aback by the challenge
from the
orthopedic surgery population.
DR. BORER: Okay.
Thank you, Peter.
Dr. Sheth, there was another
issue, I
think--
DR. FLEMING: Jeff, before we go on, could
we--
DR. BORER: Oh, I'm sorry, Tom.
DR. FLEMING: The slide came off again.
It was O-49.
DR. SHETH: Thank you.
Can we put that
back up?
DR. FLEMING: Just to make sure I'm
understanding--the denominator, actually, is keenly
of interest to me here as well, the 186
and 138.
Jeff referred to Table 18, and that was
certainly
232
one that had drawn my attention as
well. The total
MIs there when you look at on and off,
it's 144
against 113. This adds another 42 and 25 through,
I assume, a more comprehensive sweep than
got
reported here as AEs. So these are the numerators,
the 186 and 138, that correspond to the
long-term
follow-up denominators of 6,900 and
6,200. Is that
correct?
DR. SHETH: That's correct.
There were a couple other
questions I'd
like to get back to that were brought up
this
morning, and I think I just want to say
that, in
fact, not only in the clinical trial
program did we
require an ALT prior to starting; we
would require
that prior to starting drug as well to
address that
concern, and to also address the concern
of how
to--that we're going to assure
appropriate testing
compliance in practice. I'd actually like to ask
Dr. Cameron to answer that. I think that's in
response to a concern brought up earlier
by Dr.
Nissen.
DR. CAMERON: Thank you.
Hamish Cameron,
233
AstraZeneca. Really to try and help Dr. Nissen,
who posed this dilemma in terms of risk
assessment,
it's a dilemma I think we all share given
the
database we have, but recognizing it is a
very
large database based on development in
five
different indications where many of the
patients
are elderly and nearly half the patients
have renal
impairment. So it's a pretty sort of stress test
environment.
We would like to get a better
handle on
this risk. We believe that--we want to know what
the
risk of liver failure might be. We
believe
that many of the cases that we're looking
at today
are quite confounded, and it makes it
rather
difficult to reach a crisp judgment. We do not
believe the risk is as high as the 1 in
2,000
figure, but we do want to commit to
undertake work
to explore this.
Now, I think we've probably
reached the
limits of what major information we'll
get out of
clinical trial settings, and in a sort of
postmarketing
program we are developing programs to
234
exclude risks in the level of about 1 in
10,000.
We believe that would be the right sort
of gold
path to be going, and we've already
developed
programs for discussion with regulatory
authorities
both here and in Europe, and we'll be
discussing
that with FDA at the upcoming meeting.
Probably just an additional
sort of add-on
to that--and I know that, Chairman, you
have not
wanted to get into algorithms in detail
and risk
management and so on, but I think it is
an
important part of what plays on your
judgment, I
would imagine. We proposed a program based on
education, compliance enhancement, and
unrestricted
distribution. And we've read the FDA comments, and
we're obviously very eager to hear your
comments
today.
And we understand the need we have to
augment that program in order to reach
alignment
with FDA and get something in the best
interests of
patients.
And we're prepared to add more control
of mandatory elements to this program
such like
physician attestation or a patient
registry
approach, liver function testing, this
type of
235
thing, remembering that I think we have
to try and
achieve the balance, which is one of our
big
considerations with the first program,
potential
unintended consequences. Given the story
of
warfarin, given the under-treatment, we
wanted to
make sure it was a balance between doing
what's
right for individual patient safety management,
but
also thinking about the population
perspective.
We have proposals for that, and
in the
upcoming meeting with FDA, we'll be
putting that
forward as well. So I hope that gives you some
reassurance we're addressing the issue of
estimating risk to help that
judgment. We'll also
do it in a way that we get an answer
quickly as
possible through postmarketing
approaches.
DR. NISSEN: Here is what I'm trying to
weigh here a little bit. Do you know what percent
of patients in the clinical trials didn't
show up
and get their ALT done? In other words, what was
the compliance rate with getting blood
drawn for
ALT?
And I want to weigh that against what I think
we know something about what compliance
may be.
236
You understand what I'm getting at,
trying to
understand the risks.
So what percent of patients in
the
trials--do you guys know that?
DR. CAMERON: I'll let Dr. Sheth take
that.
DR. SHETH: Let's put this up. This is
overall compliance. It's a bit of a busy slide,
but it's broken up into several
components:
compliance to laboratory testing, and we
looked at
in terms of just overall monthly, weekly,
as well
as discontinuation, meaning did they meet
the
criteria and then did they discontinue on
time.
And the overall compliance was about 69
percent for
ximelagatran and comparators. The monthly
compliance was around 70, so there's not
that much
difference. And the weekly compliance overall was
68 to 53 percent on comparators. This might
reflect a little bit the SPORTIF III
trial. In
terms of discontinuing on time, about 71
percent in
the ximelagatran group.
DR. NISSEN: Now, the question, of course,
237
is:
What's that likely to look like if the drug is
brought to market? And, you know, I wonder if the
agency actually has any feedback for
me. This
would help me a lot to know. In the other drugs
where you've mandated, you know, liver
function
monitoring, do you have any sense for
what the
compliance is like in the real world?
DR. AVIGAN: Dr. Kate Gelperin has a
presentation which will deal with that
issue, I
think in some comprehensive way.
DR. NISSEN: Okay, good.
That will be
helpful.
But that was great because now I can
compare it to what we hear from the FDA.
DR. SHETH: I think the final question I
recall from this morning was did we have
some test,
a laboratory testing that might look at
the issue
when patients came off drug that might
indicate any
rebound effect, and I'd like to ask
quickly Dr.
Troy Sarich to address that.
DR. SARICH: Yes, Troy Sarich,
AstraZeneca. I can just comment in regard to the
issue of whether there's some unknown
rebound
238
effect of this agent that you want to
know about.
Everything we know about this drug is
that when
it's present in the body, it's active;
when it
leaves the body, there's no residual
effect. And
in our clinical pharmacology program, we
had over
1,100 healthy volunteers that either took
single
doses or repeated doses for up to 8
days. All of
those volunteers had laboratory testing,
coagulation assays done at the time of
discharge
from our units, and they were also brought
back 2
to 5 days after discharge. And there was really
nothing to suggest that these patients
had any
altered coagulation status after a single
dose or a
repeated dosing period. This is really sort of, to
us, a non-issue with the drug. We believe once
it's gone from the body, it's not having
any
measurable effect that we've been able to
detect.
DR. NISSEN: I hear you, but, of course,
normal volunteers are not the people who
get the
drug.
And, you know, these older people that
undergo surgery, you know, surgery
induces a lot of
very profound changes, including changes
in the
239
coagulation system. And the problem--and I think
Tom Fleming has been trying to get to
this as
well--is, you know, there's really a very
strong
signal in the data, one that actually
achieves
statistical significance, and you see it
in both of
the short-term trials, not in just one of
them, and
you see it in the pooled data and you see
it
everywhere.
So we're trying to understand
it, but it
sure looks like a signal to anybody who
would look
at this data objectively. And, you know, I know
that normal volunteer data has been done,
but there
may be effects that we don't know about,
and we're
trying to understand those.
DR. BORER: Alan, and then Bill.
DR. HIRSCH: Well, I just was going to
echo the same comment. To say that you're
satisfied is one thing, but I was looking
for the
same kind of data Bev was asking for,
which is
something very specific in a relevant
population,
thrombin gene expression rates, platelet
Factor IV,
beta thromboglobulin, d-dimers, some real
evidence
240
there was a real effort to look at
upregulation of
the pathway, and I haven't seen that yet.
DR. HIATT: Yes, just to follow up on
that, I'm reminded of the oral 2B3A
antagonist
experience, and the issue was that these
drugs has
short receptor occupancy time and that
platelets
became prothrombotic when the drug came
off the
receptor, a similar kind of PK sort of
behavior as
you have, and that the concept was when
patients
fell below inhibition of platelet
aggregation of
maybe 80 percent, they became
prothrombotic.
Now, if you look clinically, if
the
all-cause mortality rates are the same
amongst
groups, but you're preventing major
thrombotic
events, there must be some offsetting
competing
mortality weighing against that. And so I think
this question keeps coming up about is
there a
rebound, is there prothrombotic events,
and I don't
think you really know the answer to that
question.
DR. SHETH: If I can ask Dr. Troy Sarich
to respond again, thanks.
DR. SARICH: If we could just look at CP-6
241
one more time. This was from my core presentation,
and I'll just take a little more time to
go over it
this time.
We have looked at things like
beta thrombo-
globulin, which is a measure of platelet
function, and in this analysis what we
did is we
looked at a range of concentrations of
melagatran,
and plotted here is the reduction in beta
thromboglobulin levels as concentrations
increase,
and this is actually a conglomeration of
data
collected at different time points after
drug
administration, including data at 10
hours after
drug administration, which in healthy
volunteers
constitutes quite low levels of
melagatran.
You can see that as
concentrations go up,
the effect increases. When concentrations start to
go back toward baseline levels, the
effect on beta
thromboglobulin goes back to baseline.
We also see that here with
thrombin
generation using thrombin/antithrombin
complexes.
So, really, we don't see any type of
hysteresis,
any unrelated pharmacologic effects once
the drug
242
is eliminated from the body. And I take the
points, of course, we cannot exclude
anything.
There are unknowns possibly. But from all the data
we know, we have not seen any delayed
pharmacological activity beyond when the
drug is
gone from the body.
DR. HIRSCH: But I did see that slide, and
I'm satisfied that it shows that effect
in a
healthy volunteer in a non-post-operative
state
where there aren't acute phase reactants,
there
aren't external stresses. That's not really what
I'm asking.
DR. BORER: Jonathan?
DR. SACKNER-BERNSTEIN: As a follow-up to
that, even if you're measuring these
tests within
10 hours of dosing, you're still at a
point where
the drug concentration is above 0.05
micromolar,
which means you're still at a level where
you have
a therapeutic effect. And all of the questions, I
think, are focusing on what happens to
the patients
once the drug concentration is below a
therapeutic
level, but within that period of time
right after
243
it's been used.
I'd also like to propose just
another way
of trying to interpret this risk of
myocardial
infarctions and acute coronary events and
coronary
disease AEs. We're using the ESTEEM trial to say
that we're looking at a high-risk
population, and
that would be the place where it would be
the most
easy for us to detect a signal, if there
were one
there, of an adverse effect of the
drug. Although
I can't tell with certainty from the
documents
because you don't have all of the
concomitant
medications listed for all the trials,
from the
information in this I would bet that the
ESTEEM
trial patients were actually treated
extremely well
for coronary risks. So they were treated with
statins, with beta blockers, with ACE
inhibitors
and anti-platelet agents, et cetera, all
of which
may actually be protective against some
effect of
ximel--I can't say it either. You know, if indeed
there were a pharmacologic effect,
perhaps those
drugs are protective. It could also explain why
the signal for myocardial infarction
appears to be
244
weaker in the SPORTIF trials where almost
half the
patients were on statins, over half were
on ACE
inhibitors or ARBs. A large proportion of the
patients were on beta blockers compared
to the
THRIVE and the EXULT patients, where I
would doubt,
given their age and their list of
comorbidities,
that they're on many cardio-protective
meds.
So I'm not quite convinced that
ESTEEM
means that we don't have to worry about
this signal
of coronary events.
DR. BORER: Okay.
Do you have anything
else that you wanted to say, Dr. Sheth,
or are we--
DR. SHETH: I think we've addressed all
the questions. I just want to clarify just one
piece of information on discontinuations
due to
bleeding being higher on ximelagatran
than warfarin
that was presented or on comparators that
was
presented earlier. That was primarily--actually,
only driven by the SPORTIF III trial
where in the
open-label, physicians had the
alternative to put
their patients on warfarin if they came
off
ximelagatran. But if they were on warfarin, they
245
did not have an alternative if the
patient bled.
In the other trials, you do not
see the
same increase for discontinuation for
bleeding, so
I just wanted to clarify that piece of
information.
DR. BORER: Okay.
Thank you very much.
Now, let's go back to
questions, and,
again, I'm going to--I know this is
difficult to
do, but let's try to limit it to
questions of
clarification, if we can, for Dr. He and
for
Jonathan Halperin regarding the
risk-to-benefit
relation.
Anyone on the committee who has
questions, now's the time. I have a few if you
don't.
It's primarily regarding the
FDA
presentation. Starting with Slide 3, Point 2 of
the FDA's Slide 3 is that efficacy result
in Exanta
is driven by decrease in asymptomatic
distal DVT,
which is not clinically meaningful. I think this
is an important point, and I asked you
some
questions about that earlier. But I would like to
hear something else. In Jonathan Halperin's
presentation, he pointed out that 70
percent of
246
people untreated would have distal DVT,
and there
is some concern about propagation,
although it
seems that the data are not as well defined
as they
might be.
If indeed asymptomatic distal
DVT is not
clinically meaningful, that would be very
important. I'm not sure on what basis we say that.
So I'd like to hear that, and I really
want to hear
from Bill and from Alan as well about
that.
DR. HIATT: I don't think we have a real
lot of certainty about that, but the risk
is
certainly there. And it was also mentioned in the
sponsor presentation about post- (?)-itic
syndrome,
which is a very disabling manifestation
that does
occur 2 to 4 years after an acute venous
event, and
not also well characterized. So I think that risk
has to be considered as well as the risk
of an
embolus.
So I would not discount these
asymptomatic distal clots.
DR. BORER: Dr. He, on what basis do you
say it's not clinically meaningful?
DR. HE: In the clinical practice, all of
247
this data was detected by the
venogram. In the
clinical practice, usually we don't use a
venogram
to detect asymptomatic patients. If patients do
not have a symptom, you don't want to use
a
venogram to detect a patient who has a
DVT or not.
Secondly, if you have
asymptomatic distal
DVT, majority of physicians do not treat
it, and
this patient--it is basically the one you
detect by
the venogram, and in that case the basis
I say if
you found asymptomatic distal DVT by
venogram, that
is clinically meaningful, it is really--I
really
don't think it is very clinically
important.
DR. BORER: You know, I'm sorry we don't
have an orthopedic surgeon here on the
panel or
somewhere standing close by. But, you know, just
to nail down the point, because I do
think it's
important, and I want to hear from Alan
and from
Beverly about this, of course. You know, many
orthopedic surgeons would not--as you
say, they
wouldn't treat somebody because the
person had a
distal DVT. They would have treated him
beforehand. They would have given an anticoagulant
248
of some sort as early as possible after
an
operation because of the concern about
DVT. So
going and doing a venogram would be
irrelevant.
Presumably, you've already taken the
preventive
steps.
I'm still a little unsure about this.
Maybe we can go to Alan, then Beverly,
then Ron.
DR. HIRSCH: Well, I think it's a large
statement to say it's not clinically
relevant, so I
would disagree with that statement. But, on the
other hand, our database is incomplete
and we have
to acknowledge that. My instinct personally is
that it is clinically relevant, but let's
define
what that means.
First of all, we don't have long-term,
prospective, randomized trials of
treatment of
distal, meaning infrapopliteal DVT to
know its real
impact on human health in this country,
which is
crazy.
But in the absence of that, most of us
would believe that there are three
clinically
possible impacts. Just to state them, I guess not
measured in this study, you know, one
would be
propagation so it becomes clinically
relevant at a
249
small rate. The second is the quality of life
post-thrombotic impact. Nobody likes a swollen
leg, estimated to occur in probably
between 5 and
30 percent of the population, and local
discomfort.
And, more important, again, if we took a
5- or
10-year time frame long-term recurrence
of deep
vein thrombosis distal to that fact, it's
probably
pathophysiologically relevant; we just
don't know
the rates.
So I'd just be careful with not
making too
strong a statement.
DR. BORER: Beverly, and then Tom. I said
Ron, but I meant Tom.
DR. LORELL: Thank you.
And, Dr. Hirsch,
I appreciate your comments on this.
I think my comments are very
similar to
Dr. Hirsch in that the setting of this as
a
clinical trial looking at the question of
whether a
new molecule is effective in reducing
thrombosis,
as
a clinician I would look at this as a continuum
as opposed to being separate buckets of
thromboembolic events. So I think with Dr. Hirsch,
250
I would share the concern that there is
clinical
meaning to having distal venogram
detected
thrombus.
DR. BORER: We run very close to the
always difficult issue of surrogates
versus
clinically relevant endpoints here. But, you know,
I too feel very concerned about saying
it's not
important.
DR. HIRSCH: In a sense, Jeff, this is a
surrogate for the purpose of this study,
with the
possible long-term benefit if there were
longer-term data.
DR. PICKERING: The American College of
Chest Physicians, actually a consensus,
did comment
on that, and they suggest--in terms of
endpoints,
we suggest a middle ground based on large
trials
that use a clinically important VTE
outcome
consisting of a composite of fatal PE,
symptomatic,
proven DVT or PE and asymptomatic
proximal DVT.
DR. BORER: John?
DR. TEERLINK: So taking into account that
the distal DVT has some clinical meaning,
I'm just
251
trying to put it into a context of the
clinical
trial.
I would guess that the at least short-term
symptomatic, the painful thrombosis would
have been
picked up as symptomatic thrombosis in
this trial.
So that actually is accounted for in the
symptomatic group. I would expect that the
progression of this thrombus, this distal
thrombus
would have been also detected as a
proximal
thrombosis. So that part of the endpoint, at least
acutely, recognizing that it can
propagate later,
but at least within the scope of this
trial is
being accounted for in the endpoint. So I think
actually we can acknowledge that the
distal
thrombosis is important clinically, but
much of its
clinical import is being picked up by the
progression of disease in the proximal
DVT
measurements or in the symptomatic
measurements,
neither of which were different in the
study.
DR. BORER: Tom, and then Steve.
DR. FLEMING: We've got a hierarchy, as we
often do in clinical trials, of what I
would call
clinically relevant events moving on down
to what
252
would be surrogate events. And in this trial, in
this setting, you've got death, you've
got stroke,
you've got MI. Those are all very profound.
You've got PE. Then you've got symptomatic DVT.
And then in the asymptomatics, you're
down to
proximal, and Tom's list left off distal
asymptomatic, left off completely. I've heard
nothing to argue that asymptomatic DVT,
in
particular distal, is anything but, at
best, a
surrogate endpoint. It is not a clinical efficacy
endpoint.
A clinical efficacy endpoint by
definition is one that unequivocally
reflects
tangible benefit to patients,
prolongation of
survival, symptomatic disease.
If you're doing a venogram and
you
identify a distal asymptomatic, it might
mean
something. But listening to Alan, we don't know.
That's an unvalidated surrogate at
best. I don't
see that we've said anything that could
put it into
any definitional category other than an
unvalidated
surrogate. And what's problematic now is when that
component contributes the vast majority
of the
253
events and almost the entirety of the
signal.
DR. HE: I have one thing I want to point
out.
This is a short-term treatment, average
patient only using 8 days'
treatment. That means
if you detect a distal asymptomatic DVT,
you only
treat for 8 days. That means you are not waiting
for this distal DVT to progress to a
proximal DVT
because you already stop treatment after
8 days.
And, you know, this short-term treatment
doesn't
help that DVT at all.
DR. NISSEN: Yes, I think I understand the
spirit of your "not clinically
meaningful"
statement, and let me see if I can
characterize it.
I think what he's really saying, which is
really
very much what John said as well, is that
if you
didn't do the venogram, you'd never
know. And so
it's an event that you only detect
because you're
doing a procedure that would never--would
not be
done clinically in an asymptomatic
patient. And so
you're adding a bunch of events to the
mix that
would be undetected.
In other words, if you took
1,000 patients
254
and compared these therapies, you
wouldn't see any
difference unless you went ahead and
venogram'd
them all, which is not something we do
clinically.
Now, there is one issue,
though, that I
would take issue with, and that is that
because
we're talking about a short-term trial,
we haven't
factored in the possibility that those
distal
asymptomatic events, you know, DVTs,
would 6 months
from now or a year from now result in a
clinically
important recurrence. We don't know that. And I
think we have to acknowledge that we
don't know
that, but it may suggest a potential
study, you
know, to the sponsor that would be, I
think, a
tremendous addition to the field, which
is to
detect these, and then, you know, since
we don't
know what their natural history is, to
randomize
those patients to be treated or not be
treated and
find out whether any agent long term can
prevent
the conversion of asymptomatic distal
DVTs into
symptomatic pulmonary emboli or proximal
DVTs that
are worrisome.
DR. BORER: Alan, then Tom, and then I
255
want to give the sponsor an opportunity
to respond.
Maybe Jonathan would like to say
something. Tom,
did you have something? No.
DR. HALPERIN: Thank you very much, Mr.
Chairman.
I agree with the perspective that Alan
Hirsch has stated, and I think these are
clinically
important events because of our
understanding of
the pathogenesis of the clinical
syndromes. And I
would just add one other thought, and
that is this
notion that these patients are
asymptomatic,
lacking an orthopedic surgeon to testify
for you
here today, the next best thing, perhaps
quite a
distance, is a clinical cardiologist who
takes care
of these patients. All of these patients are
symptomatic. They all have painful, swollen legs
in the first 10 days after orthopedic
surgery of
this type.
And so the problem here is not
an ambient
population walking around out there in
whom we're
doing a bunch of tests and finding
clinically
trivial events. We're taking a bunch of patients
all of whom have the symptoms of deep
venous
256
thrombosis, looking to be sure that those
symptoms
are or are not related to surgery versus
surgery
thrombosis, and considering significant
the
condition that leads to proximal venous
thrombosis
and pulmonary embolism in almost every
case in
which those two more severe syndromes
arise.
Thank you.
DR. FLEMING: Jeff, maybe I do want to
comment.
DR. BORER: Tom?
DR. FLEMING: It's a clinical event
because we understand the
pathogenesis. Wow. That
would mean that the vast majority of
surrogate
markers are then clinical events. The distinction
here between a clinical event and a
surrogate
endpoint is a clinical event unequivocally
reflects
tangible benefit. It is direct evidence of an
effect on symptoms, duration of survival,
et
cetera.
A surrogate endpoint, and in
this case an
unvalidated one, is one that provides a
potential
clue that you might be having clinical
effects in
257
the manner that Steve was talking
about. Maybe we
do here, but we're not talking not about
establishing plausibility to do a
trial. We're
standing here to decide whether the
evidence is in
hand to approve an agent. And in the absence of a
signal in the non-symptomatic distal
categories, in
the absence of a signal there, we're left
with a
hypothesis that could, in fact, be
validated in a
future trial with sufficient numbers in
follow-up
to show that this is an effect that
translates into
clinical benefit. But at this point, if we only
know there's an effect on asymptomatic
distal,
that's hypothesis generating.
DR. HALPERIN: I would agree with all of
that except for the word
"asymptomatic." It means,
as you say, the surrogate, and I accept
that
distinction that you made. But none of the
patients are asymptomatic. They all have, I
presume, all of the symptoms of deep vein
thrombosis. The problem is that the symptom
doesn't correlate entirely with the
disease in this
case.
258
DR. HIRSCH: And one more series of
bridging words. This is not a clinical clue. This
is actually a hard endpoint. The sponsor should be
congratulated, actually, for taking
patients
through a clearly defined venographic
endpoint at
no essential risk, demonstrating
clot--you know,
abolition or amelioration of clot
formation.
However, though we have a hard event, the
question
is is it a hard clinical event for which
you'd seek
drug approval, and in that sense this is
sort of
like looking with surveillance echos for
LVH, carda
duplex for IMT change, renal dipsticks
for, you
know, proteinuria and saying because
we've changed
that, the patient will feel better. Mrs. Smith
taken care of by Dr. Halperin here would
not
necessarily know the difference.
DR. BORER: Steve, with your permission,
maybe we'll hold your comment about this
until we
get to the more complete--
DR. NISSEN: I have an unrelated--
DR. BORER: Oh, you have a question.
Okay, go ahead.
259
DR. NISSEN: Because I didn't get a chance
to ask it earlier, is the SPORTIF V data
going to
be published? And if not, why not?
DR. HALPERIN: We anticipate publication,
yes.
We're still in the final stages of
negotiation with the New England Journal
of
Medicine, and if they were faster, we'd
have a more
assured answer for you.
DR. BORER: Okay.
I have another issue to
raise based on the FDA presentation, and
it's
really a question for the sponsor about
the point
that the FDA raised. Do we have data--you know,
we've been concerned in the short-term
trials that
patients were only followed for 4 to 6
weeks, and
we don't really know what might have
happened to
them afterwards. But you exposed, you know, about
15,000 people, 17,000 people to the drug.
Somewhere along the way there may be some
subgroup
that was treated short term and you
actually found
out whether they were dead or alive 6
months later.
So I want to ask whether you have
anywhere in your
database something that could be
reassuring about
260
the potential for late manifestations of
toxicity
from short-term treatment.
DR. SHETH: So you're asking do we have
something like 6-month vitality data on
our
orthopedic surgery population across both the
European and North American pools?
DR. BORER: Right, anywhere.
DR. SHETH: I don't believe we have
6-month.
I think 4 to 6 weeks was the greatest
follow-up we have. The only thing I'll tell you,
though, is that in the EXULT A trial,
which
time-wise was conducted before--right
before EXULT
B, EXULT B followed it. A lot of those sites were
the same sites, and a lot of the investigators
were
the same investigators. And the conduct of that
second trial was then over the next
year. And if
those investigators were seeing their
patients come
back either with DVT or dying, I doubt
that they
would have been participating in the
second trial,
and I'm sure we would have heard about
it. So that
we had contact with all of those
investigators
during that second trial, so it's not a
formal
261
follow-up, but I would anticipate we
would have
heard something if they thought that
their patients
were having adverse events delayed.
DR. BORER: Okay.
A simple question here.
On Case 2 that was presented to us by the
FDA
showing that a major problem occurred
when a 0.25
micromolar concentration of melagatran
was present
in the blood, was the drug stopped when
the ALT was
found to be 4.5 times normal? Was it stopped ever?
I mean, there's no notation here of
stopping the
drug.
DR. SHETH: That particular patient, which
was the 77-year-old male who had an
elevation of
his ALT to 4.5 times the upper limit of
normal and
then missed two subsequent tests, at that
time the
discontinuation criteria--actually, this
is the
first algorithm that he should get weekly
testing
after 3 times the upper limit of normal,
and then
if it reached 7 times, he'd stop, or if
it was
persistent. So, no, he didn't stop his
anticoagulation. His last dose then would have
been the day--essentially before he got
admitted,
262
because he came in around 3:00 in the
morning and
the blood work was drawn then. So we don't know
what time the night before he might have
taken his
last dose, but it would have been within
clearly a
12-hour period, but we don't know.
DR. BORER: Okay.
I mean, that
uncertainty may be important in
interpreting this
blood level data, but hearing that this
horrible
situation had developed with a blood
level that is
putatively therapeutic and probably not
toxic is a
concern, and it suggests the potential
utility of
something that probably hasn't been done
because
I'm not entirely sure how you do it but
Alan
mentioned it, which is looking for
genotypic
evidence of abnormal handling or abnormal
response.
But, you know, I wouldn't have expected
you to do
it because I don't know what you'd be
looking for
yet.
Slide No. 15.
DR. SHETH: In terms of the genotypic, we
are actually conducting a pharmacogenomic
study
looking retrospectively, actually trying
to
263
understand people at risk for severe
liver injury,
and
we'll also carry forward from that study--for
any future studies our goal is actually
to
implement similar testing more
extensively in all
future trials. So we've already initiated that
effort.
DR. BORER: Well, I congratulate you. I
think that's the most appropriate thing
to do.
My next question is with regard
to FDA's
Slide 15.
This is long-term exposure, but there
were several long-term exposure
populations, and I
want to know which populations were and
were not
involved here, included here. Does this--I guess
it does with the numbers you show, but
does this
include ESTEEM and SPORTIF?
DR. HE: This is long-term exposed sick
population include all of the patients
exposed to
Exanta more than 35 days.
DR. BORER: So this is everyone who--
DR. HE: Long-term exposed sick
population.
DR. BORER: Let me just look at that for
264
one minute.
DR. HE: This one is long-term exposed
safety population. That includes all of the
patients exposed to Exanta more than 35
days. And
this table was summarized from the
sponsor's table
in Module 2, Table 53.
DR. BORER: Okay.
Thank you.
And I think I had a similar
question on
16.
As I look through the cardiovascular events
that were being tabulated, they included
myocardial
infarction and a lot of other
things. There was
angina, which I assume means any anginal
event or
perhaps it's a chest pain event. I mean, what's
angina?
DR. SHETH: The data shown on the previous
table and this table reflect
investigator-reported
adverse event, so that these were not
prespecified.
And so the physician may have just said
patient had
chest pain and that might have coded to
the adverse
event term of angina or angina
aggravated, terms
that you're seeing in those
listings. So it's just
simply investigator-reported adverse
events as per
265
their terms.
DR. BORER: The reason I'm asking this,
I'm trying to get a handle on the reality
of this
apparent cardiovascular risk associated
with the
use of this drug that we inferred from
the
short-term study and the long-term
prevention
trials, because when I look at ESTEEM,
page 6 of
the ESTEEM tab in the sponsor's book, and
when I
look at the FDA--this is the Cardio/Renal
review,
page 76, these were the SPORTIF patients,
I look at
these two trials and the results look
different
from what we're inferring and what was
admittedly
unexpected in the short-term trial and in
the
prevention, the VTE prevention trial.
On page 76, as I look through this, it
looks to me as if there are--if you
aggregate these
events, it looks as if there are fewer
cardiovascular events in the patients in
the
SPORTIF trials who were on ximelagatran
than on
warfarin.
Maybe I'm reading this wrong, but--
DR. HE: In the patients in the atrial
fibrillation population, they do show
numerically
266
higher CAD and MI events in the
ximelagatran group
compared to in the warfarin group. That is 7
percent compared to 6.7 percent. As I mentioned
before, because this is high baseline
event rate,
it's very difficult to assess the small
portion of
difference in this population, yes, but
they still
show the trend in the ximelagatran group
higher
than comparator group.
DR. BORER: Okay.
Now, you just
highlighted the question I actually had,
which I
had forgotten, so thank you. You said that this
was a high baseline risk or a high
baseline
frequency of events. And, again, I'm not
sure--just for purposes of clarification,
I'm not
sure on what basis we say that. This was a
long-term trial. The absolute frequency of
myocardial infarction was relatively low,
not
terribly unexpected given the population
that was
studied.
And angina is angina. It wouldn't
be
terribly uncommon to see angina in this
population.
So I'm not sure why we would
say that it's
a particularly high event rate. Compared to what?
267
T4B DR. HE: Compared to the
population for short-term surgical
population and
VTE population.
DR. BORER: Okay.
DR. HE: Because the difference between
groups are small, less than 1
percent. If you
assess ACS population, the baseline event
rate is
27 percent. If you assess 1-percent difference
between the group at this high baseline
rate, it is
difficult to make any conclusions. That is why I
focused on surgical population. All of the
surgery, the patient went through total
knee
replacement surgery, before they go to
surgery they
go to you, go to the cardiologist to
eliminate any
high possibility to have an MI. So that is why
that population had event rate low. In this
population, so I can see the difference,
you know,
even smaller difference, I can see
it. Same thing
for the VTE population. Baseline event, very low,
0.16.
And for the treatment, VTE treatment, only
0.1 percent. In this low event rate population, I
can see the difference. If any difference, I can
268
make an assessment maybe this difference
between
the treatment, you know, this gave me a
signal as a
medical officer, when I see this kind of
event
rate, you know, that gave me a signal,
gave me
attention--to pay attention to this kind
of thing.
So high event rate, what I mean is
between the
difference populations.
DR. TEERLINK: But, Jeff, that doesn't
make any sense. I mean, you have to have a lot of
events.
The confidence interval around these small
event rates has got to be huge. The confidence
interval around the big events rates has
got to be
small.
I totally disagree. It's the
opposite of
what you're saying.
[Pause.]
DR. HE: No.
[Laughter.]
DR. HE:
Okay, let me say it this way:
Because the baseline event rate like
we're looking
for atrial fibrillation, the difference
right now
is 7 compared to 6.7 percent. Do we know this
difference, 0.3 difference, does that
come from--
269
DR. TEERLINK: What's the confidence
interval around the difference? That's the point
I'm trying to make. The smaller the event rate,
the
bigger the confidence interval around that
difference. The higher the event rate, usually the
lower the confidence interval.
DR. HE: Yes, that is why when you go
to--okay.
Let me put it this way. If
ximelagatran
causes MI, they will cause in a very
small
proportion of the patients. It's not like liver
toxicity.
You can see it so clearly. It's 8
percent compared to 1 percent. This difference, if
any, very small. This is a small proportion of
difference. If you assess this population who have
a baseline event rate at 7, it's
difficult.
DR. TEERLINK: I agree with that because
one of the questions will be, if you want
to prove
safety around cardiovascular events in
this knee
replacement population, it's extremely
difficult
because you going to need a huge
population studied
for a long time and a lot of events to
know if
there's anything going on. The absence of events
270
makes it very difficult to know if
there's anything
going on or not.
So to try to answer safety
around a
cardiovascular problem that this drug
might cause
in this population I think is very
difficult.
DR. HE: That's true.
I agree with you.
It's difficult to assess in the atrial
fibrillation
population because those patients have so
many
other underlying diseases--hypertension,
diabetes,
you know, so many other things. Those patients may
cause MI--may have MI secondary to other
risk
factors, not secondary to
ximelagatran. That is
why in this mixed population it's
difficult to
address a single factor that comes from
ximelagatran. That is what I tried to say.
In here, in surgical
population, surgical
population has already been evaluated by
the
cardiologist before the surgery. So those
populations--
DR. TEERLINK: You'd like to think so.
That may not be true.
DR. HE: I hope so, yes, because I get a
271
lot
of this kind of consultation working in the
hospital.
You know, they just ask me clear up this
patient because this patient wants to go
to
surgery.
Anyway, that is what I mean,
because in
here there are so many risk factors to
cause MI for
the patient, diabetes, hypertension,
coronary
artery disease, so many other
things. But in here,
their baseline rate is so low so
you--probably the
risk from ximelagatran is higher than
others.
DR. FLEMING: Can I comment, maybe--
DR. HE: You can see the difference
between the groups.
DR. FLEMING: I think what Dr. He may be
saying here is that if you have an
intervention
that, let's say, increases or induces a 1 in
200
risk of MI--that's a half a percent. If that is
occurring in the backdrop of very few
MIs, you're
going to be able--i.e., a natural history
of that
population in the absence of this
intervention has
a very low MI rate, you can pick up that
effect,
about 1 in 200 in small numbers. If it's occurring
272
in the context of a 5-percent background rate
or a
big background rate, this added half
percent is
going to be diluted. You're not going to be able
to see it.
So you're right in the sense
that if
you're arguing you're going to detect a
relative
risk increase, like a doubling, then the more
events you have, the more power you have
to pick up
a doubling. But if you're talking about an
absolute increase of a half a percent, 1
in 200
patients, then the higher the background
rate, the
more diluted that's going to be, and the
less
likely or less sensitive you're going to
have to
pick that up.
DR. HE: Thank you so much. You make my
point very clear.
[Laughter.]
DR. BORER: Can I say, I think we'll have
to let it go at that, and we can do the
evaluation
internally, and then answer the questions
with that
evaluation when we get to it, because we
have some
more to do.
273
Tom, did you have a point you
wanted to
make?
DR. PICKERING: I have a question on a
different topic, and that is, in the FDA
analysis
there are statements that in the SPORTIF
trials the
results show that ximelagatran may not be
more than
50 percent as effective as warfarin,
which would
mean that it would lower the rate by, I
guess,
about 31 percent, which is only slightly
better
than aspirin, which according to meta
analysis is
about 22 percent. So I'd like to hear what Dr.
Halperin and others have to say about
this.
DR. HALPERIN: There are a number of
difficulties in interpreting the SPORTIF
results,
partly related to the selection of a
noninferiority
margin, which was really driven by a
clinical
judgment--a clinical judgment about how
much is the
disutility of warfarin worth in terms of
event
savings.
And the Executive Steering Committee, in
consultation with others and with
reference to
previous trials, and specifically with
consideration to the risk profiles of the
patients
274
enrolled, made a judgment that that's 2
percent.
Now, I'm not prepared to
discuss a
putative placebo analysis. I'll leave that to the
statisticians, because to me the
historical nature
of that analysis makes it difficult for
me to
interpret clinically.
As a clinician, I'm interested
in seeing
that my high-risk patients have low event
rates on
treatment. If I can sustain that treatment better
with one agent, that makes that agent
more
appealing to me.
This isn't a conventional
clinical trials
horse race where you're going to do a
photo finish
and decide which wins. This is an issue of
recognizing that the big picture is how
many
patients can be kept in the saddle at the
end of
the race, and from a public health
perspective, how
many horseless riders are there, because
that's the
problem.
And in the trials, I'm not even sure, to
tell you the truth, which of the two
SPORTIF trials
brings us closer to that truth.
Of course, we respect the value
of blinded
275
trials.
I'm certainly a champion of that.
But for
the last 15 years in stroke prevention
studies, we
have debated whether they are really
closer or
further from the truth. The real world of
anticoagulation therapy may be better
reflected in
the open-label SPORTIF III trial. I don't think
we'll ever see in clinical care warfarin
delivered
as it was delivered in SPORTIF V with a
common
thromboplastin for most of the
variables. There
was much less within patient variability
and much
better patient time in therapeutic range
in SPORTIF
V than even in SPORTIF III, and SPORTIF
III was
better than almost any other trial ever
reported.
So it's difficult and it's a
judgment that
the committee will have to make, and
that's why in
the design of these trials--and I, as you
know, did
co-chair the Steering Committee for
this--we said
the best thing to do is to present
both. We've
delivered the first controlled study of
blinding
anticoagulation trials for stroke
prevention.
DR. BORER: Thank you very much, Jon.
Tom will give his opinion in
the context
276
of the questions, but before we get to
that, we
have two statisticians here representing
the
sponsor.
Perhaps one or both of you want to make a
statement about the design of this
noninferiority
trial.
DR. FISHER: I'm Lloyd Fisher, professor
emeritus at the University of
Washington. I'm a
biostatistician.
I'm going to briefly present
three slides,
and I'll address two issues just because
I prepared
myself and here it is, my 15 minutes--or
15 seconds
of fame.
But these address both the percent
preserved and also the interaction within
the
SPORTIF trials.
I first heard about that in a
meeting with
Dave DeMent, who was the statistician on
the DSMV
for the trials, and Dave said that he was
not very
impressed that there even was an
interaction. I
have to say I have to agree with him,
although by
our usual standards there is. But it is 0.02--0.02
with a 3 or 6. It's not very overwhelming in the
context of a big clinical program where
you look at
277
so many different events. So there are multiple
comparison issues.
Secondly, of course, we're
talking about
ximelagatran and placebo, but this is all
on top of
actually what I think has quite convinced
me a very
large warfarin benefit with respect to
these
endpoints, even though the FDA reviewer
said
warfarin may not be any better than
placebo. I
personally just don't really believe that
in the
clinical context of all we know.
Let me talk about relative
risk. There's
a number of ways of looking at relative
risk.
Statisticians like a log scale, and I
like plots on
a log scale, and logs of relative risk
mathematically are a lot better. Things become
normal more rapidly and so on and so
forth. But if
you're talking about measurement of
effect, I
prefer one minus the relative risk, and
the reason
is that one minus a relative risk is the
proportion
of events that are avoided by a treatment
if the
relative risk is less than one, so it
actually
refers more directly to numbers of
patients and
278
something on a logarithmic scale. And the number
of individuals benefit, the fraction
preserve then
would be the ratio of one minus the new
treatment
to placebo, which, of course, we haven't
measured
because it's an active control trial,
divided by
the presumed one minus the active control
to
placebo.
That would be the proportion preserved.
In this analysis, using the
relative risk,
I make the assumption, which is true in
most of
these things, and there's no way to avoid
it, it
does have the pitfalls of historical
controls, and
from a scientific point of view, it would
be nice,
or if it were ethical to treat say AF
patients with
placebo, but it clearly is not.
Having said that, here is a
slide--ignore
the
95-percent confidence intervals for reasons
that I will talk about in a moment. In SPORTIF
III, the estimated effect to preserve
using this
method is 71.3--pardon me. The estimated percent
preserved in SPORTIF III is 116
percent. In
SPORTIF V, it's only 78.3 percent
because, as has
been pointed out, numerically the
warfarin did
279
better.
And if you pool the studies together, as
you
know from the slides you've seen, it's about a
wash.
So the estimated effect is about 101
percent.
And I computed some confidence intervals
using something called the delta method,
but I did
some simulation and what are called the
coverage
probabilities. Does the 95-percent confidence
interval actually cover the true value
under
certain assumptions? Ninety-five percent of the
time it didn't work out very well, so I
did one
other method of evaluating things, and it
also
related to how you weight SPORTIF III and
V.
For this purpose, I took a
paper by
Rothman, Lee, Chen, Chi, Temple, and Hsu,
not
because they're all at the FDA, although
that might
make it more acceptable, but because
actually it's
really a very nice paper and it has a lot
of very
good ideas and examines things in more
depth than
any other paper I know of. But using logarithmic
scale, the mathematical properties are
very nice.
So I took weighted averages of SPORTIF
III and
SPORTIF V, and the first question I asked
280
myself--personally, I do attach quite a
bit of
weight to SPORTIF III. But I think SPORTIF V for
two reasons has more weight. It is double-blinded,
but perhaps as importantly, it's here in
North
America, so that although the AF is an
area where I
think the concomitant medication and
procedures do
not vary as much as in an acute coronary
setting,
nevertheless, there could be
differences. And then
I tried to make that quantitative. Well, if I'm
going to give some weight, what does that
mean?
And what I did was I assigned a variety
of weights,
and I said, well, what weight do I need
to put on
SPORTIF III to reject the hypothesis that
the
percent preserved--that F is the fraction
preserved. I didn't mention it on the last slide.
What weight do I need on SPORTIF III to
reject the
hypothesis that the percent preserved is
at least
50 percent of the warfarin effect? And if I weight
things roughly 3 to 1, I place 3 times as
much
weight on SPORTIF V as on SPORTIF III with
95-percent confidence you have preserved
50
percent, which is not a magical number
but it is a
281
number that the Cardio/Renal Committee
has used in
the past and there is some historical
precedent.
And then I went down and I
tried some--I
said, well, what would the mixture be to
get a p
value of 0.01, of 0.002, which isn't
quite pooling
because it's this mixture, but basically is
the
pooled data. And in addition, I did the following:
The pooled data, the lower limit of the
95-percent
confidence interval using a method that I
do have a
lot of faith in is 71 percent.
So if you take all of the
SPORTIF data,
you're 95 percent confident that at least
71
percent of the warfarin effect has been
preserved.
And with that, I'll concluded.
DR. BORER: Thank you very much, Lloyd.
Does anyone want to ask any
questions of
Lloyd?
Tom, as I said, will give his analysis
later, but, Jonathan?
DR. SACKNER-BERNSTEIN: If I understand
this correctly, does that mean that
you're saying
that this kind of statistical analysis
that you
just concluded with is one that you can
use to say
282
with a degree of confidence that you've preserved
at least 50 percent of the effect?
DR. FISHER: Correct.
But there are the
assumptions about the historical
controls, and we
only do this in a situation where we
can't
ethically use placebo. But
(?) those
assumptions, that's correct.
DR. SACKNER-BERNSTEIN: Thank you.
DR. BORER: Okay.
Let's go on to the
remainder of the FDA presentation, and I
think
we've pretty much talked the rest of the
issues to
death, so then we'll be able to get our
charge and
go on to our questions. The next presentation is
by Dr. Gelperin.
DR. GELPERIN: Good afternoon. My name is
Kate Gelperin. I'm a medical officer in the FDA
Division of Drug Risk Evaluation. Today I will
present a review of
ximelagatran-associated liver
injury and the sponsor's proposed risk
management
plan.
I will discuss our assessment
of the risk
of severe or fatal liver injury with
ximelagatran;
283
our evaluation of the sponsor's proposed
risk
management plan; and present a brief
history of
risk management with hepatotoxic drugs.
To be effective, a risk
management plan
must address specific risks and have
clear goals.
In the case of ximelagatran, the nature
of the risk
may be different for short- or long-term
exposure,
and
thus may require different approaches to risk
management.
Proposed indications for this
drug include
short-term use after total knee
replacement surgery
with intended duration of therapy last 7
to 12
days.
In contrast, intended duration of therapy
for patients with atrial fibrillation or
secondary
prevention of venous thromboembolism
would be
months to years.
For the purposes of this
analysis and
consistent with FDA practice, severe
liver injury
was defined as concurrent elevation of
total
bilirubin greater than 2 times the upper
limit of
normal within 30 days of an increase in
ALT greater
than 3 times the upper limit of normal.
284
In short-term trials with
ximelagatran, a
mild liver injury pattern was seen at the
follow-up
visit with unknown potential for a
delayed injury.
No clear signal for severe liver injury
was
observed; however, potential for a
delayed injury
pattern or for risks associated with
extended
duration of therapy in patients who
require longer
post-operative antithrombotic prophylaxis
based on
current practice guidelines were not
explored. The
potential for extension of use beyond 12
days in
some surgical patients with higher risk
of
thromboembolic complications remains a
concern.
In the long-term experience
population,
abbreviated here as LTE, mean treatment
exposure to
ximelagatran was 357 days, or roughly one
year.
Substantial risk was noted for severe
liver injury
which occurred in 1 of every 200
ximelagatran-treated patients, or one
half of 1
percent.
It is also notable that there were three
liver injury-related deaths for which
study site
investigators as well as FDA considered
that
ximelagatran caused or contributed to
fatal liver
285
injury.
In the long-term clinical
trials, 37 cases
of severe liver injury were observed
among patients
randomized to ximelagatran versus 5 in comparator
groups.
The observed relative risk of severe liver
injury with ximelagatran was 6.6 and was
statistically significant compared to
warfarin or
placebo.
The lab value cut-off used to
define
severe liver injury in this severe liver
injury is
somewhat arbitrary. Additional data cuts were
analyzed for concurrent ALT and total
bilirubin
increases, such as ALT greater than 3
times the
upper limit of normal and total bilirubin
greater
than 1.5 times the upper limit of
normal. Each of
these more conservative data cuts also
showed a
highly significant relative risk for
liver injury
with ximelagatran versus comparator.
There were a total of 66, or
roughly 1
percent, of ximelagatran-treated patients
in the
long-term pool who developed concurrent
increases
in total bilirubin greater than 1.5 times
the upper
286
limit of normal and ALT greater than 3
times the
upper limit of normal. Of these, 45 cases were
judged by the sponsor to be possibly
related to
ximelagatran treatment. There were only five
comparator cases which were considered
drug-related
by the sponsor, yielding a relative risk
of 8.1,
which was also statistically significant.
You have seen this graphic
representation
of cumulative risk of ALT elevation over
time
previously.
In 37 ximelagatran-treated
patients who
developed severe liver injury, initial
signs of
liver injury as evidenced by ALT greater
than 3
times the upper limit of normal were
noted as early
as the first week in one patient, within
two weeks
in another patient, and within 30 days
for a total
of six patients. Of these six patients with signs
of early injury, alternative causes of
liver injury
were ruled out in four cases, which the
sponsor
agreed were related to ximelagatran
treatment.
The sponsor judged that severe
liver
injury was causally related to study drug
in 19
287
ximelagatran-treated patients versus two
patients
in the comparator group who met the
definition of
severe liver injury used in this
analysis. The
estimated relative risk of drug-induced
severe
liver injury equal to 8.5 was also statistically
significant.
According to the sponsor's
analysis
submitted to FDA, 14, or 39 percent, of
the 36
patients with severe liver injury failed
to
discontinue study drug at the correct
time based on
the
monitoring algorithms. In a number of
cases of
severe liver injury, there was a rapid
rise in
serum ALT from levels that were normal or
close to
normal to high levels in less than a
30-day
interval, often with a delayed rise in
bilirubin
noted despite stopping the drug.
This slide shows the sponsor's
graphic
representation of serum ALT and total
bilirubin
levels in a patient with
ximelagatran-induced fatal
liver injury that was previously
discussed by Dr.
He.
This is the patient who had biopsy
demonstrated hepatic necrosis. The Y axis on the
288
left represents multiples of the upper
limit of
normal for ALT. The Y axis on the right depicts
the multiples of upper limit of normal
for total
bilirubin, and the X axis depicts days on
study.
Serum ALT values are graphed in
red, and
total bilirubin is graphed in blue. Horizontal red
and blue lines show the cut-off for 1
times the
upper limit of normal and 2 times the
upper limit
of normal for total bilirubin and ALT,
respectively.
The purpose of this slide is to
illustrate
progression from an ALT value on day 56
that was
around 2 times the upper limit of normal
to ALT
greater than 20 times the upper limit of
normal on
day 85, a period of only 29 days. Although
ximelagatran was appropriately stopped 3
days later
on day 88, ALT and bilirubin levels
continued to
rise, and the patient expired on day 143
with a GI
bleed and coagulopathy.
Dr. Hyman Zimmerman observed in
his
textbook on drug-induced liver disease
that
instances, even very few of them, of
transaminase
289
elevation accompanied by elevated
bilirubin, even
if obvious jaundice was not present, have
been
associated with, and have often
predicted, postmarketing
serious liver injuries, fatal or
requiring transplant). In these cases no biliary
obstruction was present. Dr. Zimmerman's
observation is termed "Hy's Law"
and has been borne
out by a number of drugs.
Dr. Zimmerman noted in his
textbook that
drug-induced hepatocellular jaundice is a
serious
lesion with mortality ranging from 10 to
50
percent.
More recent mortality estimates continue
to regard the combination of pure
hepatocellular
injury and jaundice as ominous, with
about 10 to 15
percent of patients who show such
findings as a
result of drug-induced injury going to
die or
require transplant. The explanation for this
outcome is that significant
hepatocellular injury
great enough to interfere with bilirubin
excretion
must involve a large fraction of the
liver cell
mass.
In the ximelagatran clinical
development
290
program, cases of severe and sometimes
fatal liver
injury occurred despite a transaminase
monitoring
program.
We anticipate that the frequency of
severe liver injury with ximelagatran to
be
expected after marketing for long-term
use would be
equal to or greater than that observed in
the
clinical trials. The frequency of severe liver
injury in the long-term trials was 1 in
200, or 0.5
percent.
Based on a hypothetical
scenario of
100,000 patients in the general
population exposed
to ximelagatran for a similar treatment
duration
and managed by health care providers as
seen in
long-term trials, one would expect some
500
individuals to develop severe
drug-induced liver
injury.
Of these, 50 patients--that's 10 percent
of the 500--with severe liver injury
would like
progress to fulminant liver failure,
liver
transplant, or death, according to Hy's
Law.
Using the lower boundary of
estimated
mortality risk, the projected rate of
liver
failure, transplant, or liver-associated
death with
291
ximelagatran is 10 percent of 1 in 200,
or 1 in
2,000.
Consistent with this prediction, three
deaths associated with severe liver
injury occurred
in the ximelagatran long-term clinical
development
program for a proportion of one fatal
liver injury
in 2,300 patients exposed to
ximelagatran.
Beyond standard labeling, there
are three
major categories of tools as described in
FDA's
draft guidance on Risk Minimization Action
Plans,
or RiskMAPs, that can be considered in
developing
risk management plans: targeted education and
outreach, which includes providing
educational
materials to health care professionals
and
patients; reminder systems, which may
include
informed consent or dispensing of limited
drug
supply; and the most stringent category,
performance-linked access systems, which
may
include restricted access or restricted
distribution of drug. At this time the sponsor has
proposed a risk management plan that
consists of
labeling and targeted education and
outreach.
Although the sponsor has
proposed
292
transaminase monitoring to manage the
risk of
long-term use, a specific risk management
plan for
short-term use has not been submitted,
the
assumption being that the intended
treatment
duration would not exceed 12 days.
With regard to long-term use,
the sponsor
proposes to put into labeling a
recommendation for
baseline and monthly serum transaminase
monitoring,
using the second and more stringent of
the two
algorithms implemented in long-term
trials. This
is the currently proposed monitoring
algorithm,
although I understand it may be up for
discussion.
The success of the sponsor's
proposed risk
management plan is based on an assumption
that
progression to severe liver injury can be
adequately minimized through monitoring
serum ALT
at specified intervals. However, as discussed
previously, the tempo of ALT rise from
normal to
high observed in some cases of
ximelagatran-induced
severe liver injury was rapid, making a
30-day
monitoring interval for ALT less than 2
times the
upper limit of normal potentially
problematic for
293
this drug.
In addition, the sponsor's
analyses showed
that compliance with appropriate study
drug
discontinuation triggered by monitoring
in the
clinical program was variable, and we
heard some
estimates presented this morning around
70 percent
or so.
Finally, reversibility of
injury after
drug is stopped must be considered. As I showed
earlier, in at least one well-documented
case of
fatal liver injury with ximelagatran,
stopping the
drug at the time of liver injury
recognition did
not prevent progression to a fatal
outcome.
The observed compliance with
ALT
monitoring in clinical trials reflects a
best-case
performance. In practice, one would expect lower
compliance. For this reason, we anticipate that
the rate of severe liver injury observed
postmarketing would
be similar to or higher than that
seen in long-term trials with ximelagatran.
FDA experience with drugs that
can cause
idiosyncratic liver injury has shown
that, to date,
294
there are no risk management tools that
have been
proven to prevent the risk for drugs with
a rapid
rate of progression to severe
idiosyncratic liver
injury.
One caveat may be noted: Limiting
the
usage of the drug on a population basis
has been
associated with a marked decrease in
spontaneous
reports of liver failure postmarketing in
the case
of drugs such trovafloxacin and pemoline.
Troglitazone, or Rezulin, an
oral
hypoglycemic agent, is an example of a
drug with
so-called Hy's Law cases observed during
clinical
trials portended a significant
postmarketing issue
with severe liver injury and fatal liver
failure.
In response to reports of liver failure
received by
FDA after troglitazone approval in 1997,
a series
of "Dear Health Care
Professional" letters were
sent to practicing physicians warning
about severe
liver injury and recommending monthly
transaminase
monitoring.
A study of compliance over a
3-month
period showed that only about 5 percent
of patients
received the recommended monthly
monitoring for 3
295
consecutive months in the study. Troglitazone was
withdrawn from the U.S. market in March
2000 after
94 cases of drug-induced liver failure
had been
reported.
An analysis of 94 cases of
liver failure
which had been reported to the FDA showed
that the
progression from normal hepatic function
to
irreversible liver injury occurred within
less than
a 1-month interval in 19 patients, who
were
indistinguishable clinically from the 70
patients
who had an unknown time course to
irreversibility.
Progression from jaundice to hepatic
encephalopathy, liver transplantation, or
death was
rapid, averaging 24 days. The authors concluded
that progression to irreversible liver
injury
probably occurred within a 1-mont period
in most
patients, casting doubt on the value of
monthly
monitoring of serum transaminase levels
as a means
of preventing severe drug-induced liver
injury. A
key issue in effective intervention to
prevent
fatal liver injury is recoverability at
time of
sign or symptom onset.
296
In the clinical trials which
led to
troglitazone's approval, there were no
cases of
liver failure. In the NDA database, n equaled
2,510; 1.9 percent of
troglitazone-treated patients
had ALT greater than 3 times the upper
limit of
normal, and five patients had ALT greater
than 30
times the upper limit of normal, two of
whom had
jaundice.
Although the size and extent of exposure
to study drug was very different in the
troglitazone and ximelagatran clinical
programs,
with many more patients studied long term
on
ximelagatran, some comparisons may be
made.
Unlike ximelagatran, there were
no cases
of acute liver failure or fatal liver
injury
observed prior to troglitazone
approval. Also, a
more than four-fold difference is seen in
the
percent of patients with ALT greater than
3 times
the upper limit of normal, with roughly 8
percent
on ximelagatran and 2 percent on
troglitazone.
Consideration of the markets experience
with
troglitazone may be relevant to risk
assessment of
ximelagatran.
297
Bromfenac, DURACT, which was
approved by
FDA in 1997 for use as a short-term
analgesic for
periods of 10 days or less, is an example
of a drug
which could have been used safely for
short
periods, but, unfortunately, during
marketed
experience the drug was used in excess of
the
recommended duration. No cases of liver failure or
fatal liver injury were seen in clinical
trials.
In short-term trials, the product showed
a low rate
of 0.4 percent of patients with ALT
elevations
greater than 3 times the upper limit of
normal. A
much higher rate of transaminase
elevation was
observed in patients with osteoarthritis
or
rheumatoid arthritis who were treated in
longer-term trials. For this reason, bromfenac was
approved with a warning that the
short-term
management of pain should be less than 10
days'
duration, but liver enzymes should be
monitored if
used for more than 4 weeks.
Post-approval, reports of
hepatic failure,
including four deaths and eight cases
requiring
liver transplant, were received. All but one of
298
these cases involved the use of bromfenac
for more
than 10 days, the maximum recommended
duration of
treatment. In response to the reports, FDA and the
company strengthened the warnings in the
U.S.
package insert with a black box, and the
company
issued a "Dear Health Care
Professional" letter.
Despite these efforts, the FDA and the
company
continued to receive reports of severe
injuries and
death with long-term use of
bromfenac. Given the
availability of other therapies, in 1998
FDA and
the company concluded that it would not
be
practical to implement the restrictions
necessary
to ensure the safe use, less than 10
days, of
bromfenac and the drug should be
withdrawn from the
market.
The effectiveness of
transaminase
monitoring in preventing severe
drug-induced liver
injury has not been convincingly
demonstrated.
Transaminase monitoring is ineffective
within the
tempo of liver injury is such that
inexorable
progression occurs, even after the drug
has been
stopped in response to a signal of
transaminase
299
elevation. The foremost requirement that
determined the usefulness of transaminase
monitoring in preventing frank liver
injury is at
the time interval between onset of liver
chemistry
abnormality and subsequent liver injury
must exceed
the screening interval. Rapid acceleration of
liver injury in some individuals may
preclude an
absolute protective value of standardized
periodic
transaminase monitoring.
In summary, the sponsor has
submitted a
risk management plan based on voluntary
monthly ALT
screening via product labeling. As outlined by the
sponsor, the stated objectives for this
risk
management plan are to facilitate
compliance of the
monitoring recommendations by health care
workers
and patients through education and to
minimize the
risk of severe liver injury. FDA is concerned that
it is unlikely that the risk of severe
and
potentially fatal liver injury will be
adequately
minimized by the sponsor's currently
proposed risk
management plan.
The sponsor has not
demonstrated that
300
compliance with monitoring postmarketing
would
protect patients and, even if full
compliance were
achieved, that ALT monitoring can prevent
serious
liver injury with ximelagatran.
In addition, the sponsor has
not proposed
a strategy to prevent prolonged use of
this drug
after total knee replacement surgery.
In conclusion, ximelagatran can
cause
severe and even fatal liver injury in
some
patients.
Initial signs in patients who developed
severe liver injury were noted during the
first
month of ximelagatran use in six patients
from
long-term trials. The ability of transaminase
monitoring to adequately minimize the
risk of
severe or fatal liver injury remains
unproven for
ximelagatran.
To date, serum transaminase
monitoring in
ximelagatran-treated patients has not
been
demonstrated to be effective in
preventing
idiosyncratic drug-induced liver injury.
Currently, the proposed monitoring plan
does not
provide assurance of safeguarding the
patient
301
against developing a rapid onset and
life-threatening reaction.
I would like to thank my
colleagues in the
Office of Drug Safety and recognize other
primary
collaborators on this project, including
Directors.
Allen Brinker and Claudia Karwoski, and
especially
my division director, Dr. Mark Avigan.
DR. BORER: Thank you very much, Dr.
Gelperin.
Rather than have any questions
of Dr.
Gelperin at this point, why don't we go
on to Paul
Watkins, who's sitting to my left here,
to talk
about drug-induced liver toxicity. And if we have
any clarification issues or informational
issues
for cardiologists who don't know where
the liver
is, we can do that. Oh, it's somewhere below the
heart.
Then we will move on to the charge and the
questions and have any other discussion
within the
committee itself.
Paul?
DR. WATKINS: Okay, thanks. I'd like to
thank the committee for inviting me
here. My
302
charge is to discuss some principles of
drug-induced liver injury that are
pertinent to
assessing the data that you have. I won't talk
specifically about ximelagatran in my
short
presentation, but would be happy to take
any
questions that you may have.
Just to frame the discussion,
one of the
very interesting things about drugs in
the liver
and drug-induced liver injury, or DILI,
as it is
called, is that there are many different
forms
histologically and clinically, and I've
listed some
of the many different forms of DILI that
can occur
with drugs. Drugs characteristically tend to have
a signature or a characteristic injury
they produce
which can be any one of these
patterns. But one of
the observations that's been made is the
drugs that
have entered the marketplace and been
discovered to
have more of a liver safety concern than
was
believed at the time of approval and
leading to
regulatory action are really almost--you
can almost
make a blanket statement, none of these types
of
liver injury but a specific type of liver
injury,
303
which is hepatocellular injury, and these
are the
drugs that have undergone either
withdrawal--we
heard about these two--been relegated to
second-line status, or received--or there
was a
communication, usually a "Dear
Doctor" letter or
direct-to-consumer advertising. And I won't go
into them specifically, but with the
exception of
valproic acid, an anti-seizure drug
that's been on
the market for a long time and causes
microvesicular
steatosis, and possibly terbinafine,
which has hepatocellular injury but also
accompanied usually by a cholestatic
component, and
based on my experience--and I've been
involved with
most of these drugs--the issue that got
them into
trouble was an acute hepatocellular
injury
progressing to acute liver failure.
Oh, and the other thing I
should say, with
the exception of acetaminophen, which is
a
dose-related hepatotoxin, the others
would be what
we would call idiosyncratic. So 13 of the 16 drugs
undergoing those regulatory actions in
the last 7
years, the action was because of acute
304
idiosyncratic hepatocellular injury.
Hepatocellular means the liver cell
itself is
attacked, breaks open releasing its
contents, which
includes ALT and AST.
Idiosyncratic--this is a slide
that I
borrowed from John Senior with
permission--apparently has this
derivation back in
early Greek times, with idios, one's own,
self;
syn, together; crasis, mixing or
mixture. And,
therefore, this refers to a person's own
mixture of
characteristics, factors, nature and
nurture--it
sounds like John--uniquely. The aspects that make
that individual uniquely susceptible to
the injury.
Now, the characteristics of an
idiosyncratic acute hepatocellular
injury, which,
again, has been the issue in the 13 of 16
drugs
that have undergone regulatory action,
this is a
delayed reaction, the person has no signs
of liver
problems, normal liver chemistries for
weeks or
months, and then it occurs;
characteristically has
high serum transaminases, ALT and AST,
with
generally an unremarkable alkaline phosphatase,
305
modest elevation. If the alkaline phosphatase is
markedly elevated, that is then a mixed
cholestatic-hepatocellular injury, which
is not,
with the exception of terbinafine, what
we're
talking about. Also, it's a rare event, and when
the patient becomes jaundiced, as we've
heard
about, there's a life-threatening
injury. And as
we heard, this was first pointed out by
the late,
great Hy Zimmerman, who made this
association in a
variety of drugs. I've just listed four here from
published series. These are reports of postmarketing
events.
I'm sure you can't all read
this, but the point is if you look at the
number of
deaths that occurred among patients that
were
jaundiced with these four drugs, the
average is
around 10 percent. So this is real-world, somebody
becomes jaundiced, in general not liver
chemistry
monitoring, go seek a physician, and have a
10-percent mortality. And his contribution was
saying it's sort of independent of a
drug. If it's
an hepatocellular injury, the mortality
is about 10
percent on average if you develop
jaundice.
306
Characteristics of drugs that
are capable
of causing this idiosyncratic severe
liver injury
is that in clinical trials they do almost
to a man
have an increased incidence of ALT
elevations
greater than three times the upper limits
of normal
relative to placebo. However, there are many drugs
that have ALT elevations such as
statins. Heparin
is another example, very common. So ALT elevations
in themselves are not a concern. With these same
drugs, the majority of people who have
the ALT
elevations are actually not at risk of
developing
significant liver injury even if they're
continued
on drug unmonitored.
The concept that has evolved,
if you look
at this triangle as being all patients
being
treated with these drugs, the vast
majority shown
as the bright green can take them totally
safely
with no reason to believe there's any
liver injury
whatsoever. There's a subpopulation that develops
elevated serum ALTs, yellow triangle, but
even with
continued treatment, most of those will
actually
resolve and come back to have normal
serum ALT.
307
The liver adapts, and this adaption
process is
poorly understood in the NIH research
plan for the
next five years from the NIDDK, which I
helped put
input into. Studying this adaption phenomena in
animal models is one of the high priority
items.
But not everybody adapts. A subset if
continued on drug will progress to
jaundice, and at
least in the real-world setting, 10
percent of
those will develop acute liver failure.
This is a slide from a review
that I wrote
months ago. It's not a real patient, but it's a
example of what you would expect in a
typical acute
hepatocellular injury leading to death
from a drug,
and like the other slide, what we have is
days on
drug on the X axis. This was John, Sr.'s. One of
his many contributions was deciding to
graph the
liver chemistries as a function of log of
upper
limit of normal so you can put all of
them on the
same axis.
The point here--and I'm losing
my
pointer--is that serum ALT and--ALT and
AST remain
normal on the drug for the first
month. Then a
308
little after two months they're up. At this time
the serum bilirubin is totally
normal. Drug is
continued out here beyond day 120. The serum ALT
continues to rise. Bilirubin is still normal, but
if the injury progresses, then the
bilirubin begins
to rise.
This is because the liver has lost its
functional ability, lost too many
functioning
hepatocytes.
The drug is stopped in this case, but
then
the injury progresses, and that's
actually
something characteristic that was
referred to
earlier, that once you develop this
inflammatory
response in the liver, it doesn't revert
right away
and will take a while to resolve.
In this case you can see the
serum
transaminases are coming down nicely and
might be
confused for this patient getting better,
but in
fact there's no liver left to leak
transaminases,
as indicated by the continued rise in
bilirubin.
Now, a patient may or may not
have
symptoms during this phase when the ALT
is up and
the bilirubin--before they become
jaundiced, which
309
would be in about here, in most cases
they do not,
and that's obviously the rationale of
monitoring is
to catch them at the asymptomatic phase
before they
go through and progress.
This pattern is more or less
similar in
all those drugs I showed you on the other
slide.
However, the time to onset the
characteristic rate
of upstroke, how quickly it reverses when
you stop
the drug, vary between one drug or
another. But in
general you don't in clinical trials that
I've been
involved with actually have a case of
someone who
has gone all the way through to a fatal
outcome.
Therefore, what you're left is trying to
assess
lesser signals than that. And the reason is in
general these are very rare events. You don't
treat enough people. You don't treat for long
enough.
The denominator is really people that have
taken the drug for months. Most importantly, once
you realize there may be a liver issue,
you stop
the drug in clinical trials when the
serum ALT gets
to a certain point, so you don't see the
natural
history.
310
So in evaluating liver safety
databases
the usual thing is to try to find people
at various
points.
How far did they go along in this trial?
And that's the basis of this Hy's Rule,
that if ALT
elevations alone don't mean very much,
combining
them with bilirubin elevations, even if
it's
reversible, tells you the drug has this
ability to
go all the way through.
Just ALT elevations are not very
predictive. I mention that. However, the higher
they are or if they're accompanied by
signs of
hypersensitivity, fever, rash,
eosinophilia, or
symptoms that suggest this has real
inflammation
systemic symptoms, we get concerned. 8 or 10 times
the upper limit of normal is a cut to
look at for
that.
We talked about Hy's Rule. This
is the most
conservative one with a bilirubin just
1-1/2 times
the upper limit of normal, that this is
somehow
more predictive of a signal. And so what everybody
does is they ask for all the cases where
ALT was
greater than three times and bilirubin
was greater
than 1-1/2 times, and look at those cases
very
311
carefully.
And what you look for is, is
there a cause
other than the drug, number one, but
number two, is
this a hepatocellular injury or is there
evidence
of cholestasis, elevated alk phos, in
which case
the bilirubin elevation is occurring at
an earlier
stage before there is serious liver
injury.
So first determine whether each
case is
consistent with hepatocellular injury and
that the
drug is the cause. And then what's commonly done,
once you have these selected cases after
this
analysis, is assume irreversible liver
damage will
occur in 10 percent of these cases. And of course,
that's the 10 percent rule really came
from
post-marketing observations of people who
walked
into their physician once they were
already
jaundiced. And as we heard, the only clinical
trials data that supports the
predictiveness was
with troglitazone, where 2 out of roughly
2,000
people had true hepatocellular jaundice,
and that
would predict around 1 in 10,000. That may be
about right post marketing in terms of
the liver
312
events.
The problem with extrapolating
this data
is there were no criteria for stopping
liver
chemistries in the troglitazone clinical
trials up
through NDA submission, so it was up to the
individual physician to decide whether
they wanted
to stop.
So things were allowed to progress to a
greater stage.
In addition, troglitazone
characteristically has a very long tail
of
recovery.
Even ALT elevations took greater than
two months to come back to normal, so
that stopping
the drug, you know, more likely the
progressive
liver injury. So it's difficult to extract that
data back to other drugs.
Then my last comments is--I
think they've
been made--which is the effectiveness of
monitoring
really is going to be a function of,
number one, if
patients are symptomatic. So, for instance,
Isoniazid, pulmonologists feel they can follow
symptoms reliably, and most public health
services
don't do routine liver chemistry
monitoring as a
313
safety measure for that. But then it's the rate of
upstroke, how quick does it go up? And obviously
it goes up very quickly. Your interval would have
to be very short. And then when you stop it, how
quickly do things return back to normal
are kind of
the specific issues.
So to finish up, my take-home
points are
that isolated ALT elevations are
difficult to
interpret, but if it's greater than 8 to
10 times
or associated with symptoms, it raises
concern.
The highest concern is bilirubin
elevations in the
setting of a hepatocellular injury, but I
think you
need to look at the cases carefully to
look for
high serum ALTs in the presence of a
relatively
unremarkable alk phos, so you're talking
about
jaundice due to hepatocellular
injury. And then
our ability to predict the true risk,
especially in
the real world from safety databases is
imperfect.
Thank you.
DR. BORER: Thank you very much, Paul.
We'll take a couple of minutes in case
anyone has any burning questions about
these safety
314
issues.
We've had some superb presentations here.
But I don't want to dwell on this because
we do
want to move on to our discussion.
Is that Tom?
DR. PICKERING: Yes.
Could you comment on
the statements that were made earlier,
the 16 out
of 18 patients who a rechallenge did not
develop
additional elevation of ALT, and also
that those
who maintained on the drug, it appeared
to be a
transient phenomenon. How would you interpret
that?
DR. WATKINS: The fact that you could
rechallenge people and not see a return of elevated
ALT is reassuring in the sense that that
argues
very strongly against some sort of
immunoallergic
or hypersensitivity reaction, so that
there would
be even a greater reaction the second
time with the
drug.
It doesn't, however, mean that
the
elevation wasn't due to the drug, and
there are
examples of that where drugs that clearly
have
caused ALT elevations on rechallenge have
blunted
315
or not ALT elevations. And it's presumably part of
this poorly-understood adaptive
mechanism, where
the liver actually adjusts in some way
and has a
memory or a prolonged memory effect.
And that's the answer to the
next question
you had which is why is when you continue
to treat
these people the serum ALT comes back to
normal?
And as I say, that's a fairly universal
finding.
So, for instance, with Isoniazid, can
cause acute
liver failure in some studies of 1 in
1,000 people
treated for tuberculosis. The incidence of ALT
elevations greater than three times is 15
percent.
Well, obviously, that 15 percent,
unmonitored,
don't go on and develop acute liver
failure. The
majority of them adapt, and the current
thinking is
it's a few patients who are incapable of
adapting
that go on and develop progressive liver
injury.
DR. BORER: Steve?
DR. NISSEN: Could you give us an example
where Hy's Rule fails? I mean is there out there
anything where it clearly, there were
people that
had--they had elevations of enzymes and
316
hyperbilirubinemia, where just nobody
goes and
develops liver failure?
DR. WATKINS: Not that I'm aware of, and
of course, I usually get consulted when
there
clearly is a liver safety issue.
I think the confusion is
sometimes if all
you do is look at ALT and bilirubin and
don't look
at alkaline phosphatase, you may be
picking people
that in fact have a large cholestatic
component
where jaundice occurs very early in the
course of
injury, and then you would be not
extrapolating
appropriately for the risk of
irreversible injury.
But to answer you question specifically,
I'm not
aware of any, and I think that's part of
the reason
why the Agency right now is feeling quite
confident
about extrapolating Hy's Rule to the real
world.
DR. BORER: Tom?
DR. PICKERING: Could you also comment
about the statins? Because I think when they were
first released there was concern about
this issue
which seems to have largely gone
away. And also,
overall in the ximelagatran groups there
was no
317
significantly higher bilirubin levels
than in the
comparitor groups. Is that of any relevance?
DR. WATKINS: Well, let me answer the
first question. In terms of statins, the concern
was there was a preclinical model that
clearly
developed serious liver disease, and
those drugs I
showed you, in general the preclinical
studies have
been completely clean. So going into the early
statin trials there was already a concern
about
liver, severe liver injury, and then
seeing the ALT
elevations carried that concern
forward. I think
it's pretty clear the risk is quite small
now from
the statins as a group, and for reasons
that aren't
totally clear. But there are ALT elevations, as I
point out, that don't predict subsequent
liver
injury.
Now, what I don't know is in
the clinical
trials of the statins whether there were
any Hy's
Rule cases. That would be a very interesting
question to explore if you just used a
very low
cutoff like ALT 3 times, bili 1-1/2
times, and I'm
sure that's being looked at at current
statins, but
318
in the past, I'm unaware of it.
DR. BORER: Okay.
Thank you very much,
Paul.
I think what we'll do now is go
on to the
questions, and we'll have the rest of our
discussion in the context of the
structured
questions before which we need to hear
what we're
asked to do. So if we can have a statement about
the charge to the Committee from Dr.
Korvick.
DR. KORVICK: Thank you, Mr. Chairman.
I would just like to point out
that for
your convenience, we have provided a hard
copy of
the questions and also we're prepared to
project
those questions on the screen for
everyone to see.
In the interest of time I'm not
going to
read the questions to you, but point out
the areas
of interest. As you will see when you read through
these, we are very interested in the
Committee's
opinion about the safety and the data
that you've
heard presented today. Then we go on to ask you
about your opinion on the benefit risk in
the three
proposed indications that are presented
for you.
319
I would just like to point out
that the
Committee has been constituted in such a
way with
the expertise to be able to appropriately
address
for us this issue of benefit risk
evaluation, and
that is very important for us to
understand. As
you've heard earlier today, the FDA and
AstraZeneca
will be meeting and having further
discussions on
the risk management program. However, based on
your concerns about the benefit risk and
these
various indications, it is conceivable
that further
advice may need to be sought from the FDA
Drug
Safety Committee if it comes to that.
So I think we are very
interested in your
perceptions on the benefit risk, and we
are also
interested that you ask the Committee
members to
vote on all the questions.
Thank you.
DR. BORER: Thank you, Dr. Korvick. I can
assure you, you will hear our opinions.
We have two non-voting members on
the
Committee, Dr. Vega and Dr. Watkins, who
can
participate in the discussion but not
vote. But
320
I'm going to ask for a vote as well as an
opinion
if you choose to give one on all the
issues listed
in these questions. We have a Committee reviewer,
Steve Nissen, and a Committee statistical
reviewer,
Tom Fleming, and for some of these issues
I'm going
to ask them to go first, and then go
around the
table, and in others we'll just go around
the
table.
On the first question regarding
safety,
what we want is a statement about level
of concern
from everyone, none, low, moderate or
high, for the
risk of liver toxicity with the use of
ximelagatran
in each of the three settings that are
noted here,
and we'll want an explanation for each of
these.
Steve, why don't you start out
here?
MR. NISSEN: In my answer, I will lump
1(a) and 1(b) together because they both
relate to
the longer term use of the agent.
I would consider the risk to be
high, and
the reasons are that by Hy's Rule, one
would
estimate approximately a 1 in 2,000 risk
of acute
liver failure, fulminant liver
failure. By actual
321
observation it's 1 in 2,300. So we have two
independent sort of sources of
estimation, both of
which give about the same estimate. In addition,
I'm troubled by the fulminant nature of
the liver
injury that is seen, that is difficult to
predict.
Even with monthly monitoring, in 29 days
a patient
can go from first elevation to
irreversible fatal
injury.
My judgment from hearing the
three cases
is that they are all three almost
certainly drug
related, and so again, I can't make that
problem go
away with the chronic use, so I do think
the risk
is very high.
One of the things I did look at
as I
reviewed this is what the rate was for
another
agent that was withdrawn, which is
troglitazone,
and using this methodology, you get about
500 per
million patient years for ximelagatran,
and that is
just about double the rate that was
observed with
troglitazone. So I think based upon the available
data, the estimated risk here exceeds
that of an
agent that was withdrawn for these kinds
of safety
322
considerations. Now, that says nothing about
benefit, but that's my view of absolute
risk.
With regard to short-term use,
my
assessment of the risk would be lower,
but I am
worried about a couple of things that I
suspect the
Agency's also worried about. We hear a lot about
dose creep when we talk about drugs for
arthritis
and so on. I'm worried about duration creep here,
that, you know, saying to somebody,
"You can use
this drug for 12 days, but not 13, 14, 15
or
longer," and I know my colleagues and
I know how
they think, and my concern would be
somebody's on
it, they seem to be doing well, they have
prolonged
risk of venous thromboembolism, and so
they stay on
it for 20 or 30 days maybe, not six
months.
We don't have a lot of data on
the delayed
risk profile, what happens if you give
the drug for
say three weeks or two weeks and then
stop it, you
know, it there a late phenomenon?
So my risk concerns are lower
for
short-term use but they don't go away in
short-term
use because I'm concerned about this
duration creep
323
problem.
DR. BORER: Tom, do you want to address
that question?
DR. FLEMING: The only thing that I might
add which certainly struck me as
particularly
noteworthy is in the long-term database
with the
6,900 and 6,200 that when we look at Hy's
Law and
we look at these people who are
categorized as
severe liver injury, there's a half a
percent
increase as has been repeatedly
noted. And if Hy's
Law applies, that is in fact then 1 in
2,000 that
would progress to liver failure, transplant
or
death, which is certainly very
significant.
The other thing that I noted
was in the
third indication listed here in 1(c),
what is
certainly, what caught my eye was that
it's not as
clear to me what the risk is, but it's
interesting
in the 4- to 6-week period at 7 to 1, and
I would
sure like to know what happens after week
6 here.
DR. BORER: So we need a precise answer,
so if you are agreeing with Steve, that
means high
for (a) and (b), and what for (c)? Tom?
324
DR. FLEMING: What is my answer for (c)?
I mean I agree high on (a) and (b), and
(c), I
don't know, it's unknown, but I'm
concerned.
DR. BORER: Why don't we start at that end
of the table with Tom?
DR. PICKERING: I think I would say
moderate for (a) and (b). A lot of this depends on
the fate of three patients out of nearly
7,000, and
of the 37 with "severe liver
injury," most of them
recovered or died from unrelated causes,
and of the
3 cases, my interpretation was that one
died from
hepatitis B, one probably from drug related
and the
other one was a mixture. So I think I wouldn't
rate it as high as Steve. It's moderate.
DR. BORER: And what about (c)?
DR. PICKERING: (C) I would say low,
assuming that it's given for the stated
duration.
DR. BORER: I'm sorry.
Dr. Vega, you're
not a voting member, but if you have any
major
concerns that you want to raise here,
please go
ahead.
DR. VEGA: I think given the data I would
325
say that I would have a moderate level of
concern
for (a) and (b) and low for (c).
DR. BORER: Ron?
DR. PORTMAN: I would have to agree that
(a) and (b) would be high. I think we don't
understand very well the injury, the
liver injury,
and I think that for me, whatever plan
for
monitoring the company may have, that we
will want
to see a study to see if that were in
fact
effective. As far as (c) is concerned, I would
rate that low.
DR. BORER: Bill?
DR. HIATT: Pretty much the same
conclusions. I think for (a) and (b) it's high.
It seems that we have more evidence from
this data
than other studies about progression of
liver
disease on to serious consequences. So though the
absolute risks are low in the actual
clinical
trials data, the real risk to the
population I
think would have to be considered high.
I think for the knee surgery
population,
it's unknown, but probably moderate at
this stage,
326
and easily quantifiable over a relatively
short
period of time, you know, 6 months. So my vote
there would be moderate.
DR. BORER: Beverly?
DR. LORELL: I think for the two long-term
groups for the reasons that have been
stated the
risk is high. I think for the surgical group as
was used in the trial, the risk is low.
DR. BORER: Jonathan?
DR. SACKNER-BERNSTEIN: I would say that
even though the sponsor has made the
statement that
they think that they can reduce the risk
to 1 in
10,000 instead of the 1 in 2,000, even if
that were
the case, I still think as you do the
math, that
leaves you still at a high risk in
letters (a) and
(b).
And for (c), one of the factors
that we
didn't even discuss was the fact that
several
patients, after discontinuation of drug,
developed
ALT abnormalities. Since the follow-up was only
four to six weeks, and it's been
established pretty
clearly that the peak rise in these liver
enzymes
327
is between two and six months, I don't
see how we
allow the absence of safety data to allow
us to say
that it's no or low risk. I think at the minimum
you'd have to say it's a moderate risk,
so I'd say
moderate for (c).
DR. BORER: Susanna?
DR. CUNNINGHAM: I would also vote high
for (a) and (b), and for (c) I would also
vote
moderate, partly because of the FDA data
that Dr.
Gelperin presented that the other drugs,
people
have not kept the dosage duration down,
so I think
it's likely to creep as Steve said.
DR. BORER: Paul, you're not a voting
member here, but again, if you have any
opinion to
give about this, we'd love to hear it.
DR. WATKINS: Just in regards to (c), the
concern that a few patients had elevated
transaminases a few weeks after stopping
short-term
treatment. I think it would be unprecedented for a
drug that causes hepatocellular injury or
any of
the drugs that I listed that underwent
regulatory
actions in short term to weeks later
developed a
328
problem, that is seen in a few drugs that
have
cholestatic features like augmentin,
where you can
stop the drug and weeks later there can
be a
problem.
And even then, that would not generally
be thought of as a major medical
problem. So it
would be helpful to actually see the data
on those
patients to be sure, but I would be very
surprised
if any subsequent studies showed that a
short-term
treatment led to significant liver
problems weeks
or months later. It would be unprecedented for
hepatocellular injury in my experience.
DR. BORER: Thank you.
I would vote high for (a) and
(b), but I
have to provide a caveat and
explanation. It's
hard for me to talk about safety without
putting it
in the context of the relation to
benefit. I'll
say right now, and it will come up I'm
sure from
others as we continue this discussion,
that
warfarin is a very difficult drug to
use. There
are many, many problems with its
use. There are
situations and patients in whom it's not
practical
to consider using it. Having an alternative that
329
can really markedly reduce risk in
certain
situations is something that's very
attractive to
me.
Jonathan Halperin showed a risk to
benefit
relation slide that was broad strokes,
but still
suggested there is a population, if
Coumadin were
not really a practical option, that
certainly would
benefit more than it would be at risk,
the risk of
hepatocellular death worse case or liver
failure
death at worse case we're sort of
currently
estimating at 1 in 2,000, although that's
a total
guess because there are so many unknowns
here. And
the number needed to treat to prevent a
major event
is 1 in 10 or 1 in 15.
The risk in absolute terms is
high
relative to what I consider to be the
risks of
other drugs that I use in patients with
cardiac
diseases, but you do have to think of it
in terms
of the relation to benefit.
Nonetheless, I think that the
risks of
hepatic injury, of liver toxicity with
ximelagatran
is high in the setting in which it's
going to be
330
used for a long time. Whether that can be
minimized with a monitoring program
remains to be
determined. I don't know that. I think that's
something that the FDA is going to have
to talk
with the company about, and I applaud the
potential
use of genetic profiling, because as Paul
pointed
out, it sort of sounds like this is
probably not
dose related, but be idiosyncratic. And if it is,
then you have to find out the
characteristics that
identify the people at risk.
For the prevention of VTE in
patients
undergoing elective total knee
replacement surgery,
I have a relatively low level of concern
about
safety if the drug were used as it was
used in the
trial.
But as several people have said now, I
can't believe that it will be, and
therefore, we're
going to have an overlap between
short-term use and
long-term use, and once you get into
long-term use
or longer-term us, you do begin to worry
again
about the unknown risks and the potential
remediation of those risks by a
monitoring
algorithm that still has to be
determined.
331
So that's a long-winded answer,
but I said
high, high and sort of low.
John?
DR. TEERLINK: So I'll say high, high, and
the thing that reinforces that for me is
the
consistency of the data, and the other
thing we
hadn't talked about was even if you
forget about
the ALT plus bilirubin, there was the ALT
greater
than 10 times the upper limit normal
being 15-fold
higher in the ximelagatran group, so that
just
reinforces for me the concern in that
group.
Prior to what Paul said I had a
moderate
concern about the post total knee
replacement
group, the short-term group. When you have an
8-fold or greater increase in--8-fold
greater
number of patients who have had ALT
elevations and
no follow-up data to know what happened
to those, I
was going to say moderate just to help
make sure we
get that information. If we are confident that
those kind of short-term exposures don't
tend to
result in longer term deficits, then I'd
be
comfortable saying low.
332
DR. BORER: Dr. Sjogren.
DR. SJOGREN: I think for (a) and (b) I do
agree that the risk is high. However, here's the
dilemma:
it's high for a small proportion of
patients because it is idiosyncratic in
nature.
Therefore, I hear this drug is good
otherwise, you
know, probably preferable to Coumadin in
many
patients, and so when we think about it,
we think
about the risk, but then we think about
the benefit
for the other patients that won't develop
the
fulminant hepatic failure.
So I think although I rate it
high for a
small number of patients, still I think
it would be
very important for us to find a way on
how to
monitor so we can prevent the very bad
outcomes,
such as transplantation or death.
For (c) I think the risk is low
considering what we have discussed.
DR. BORER: Alan?
DR. HIRSCH: It's hard to be at the end of
the series and say anything novel, but I
will
ignore the benefit which I think does
probably
333
exist for this medication, remembering
that
Coumadin does have high rates of adverse
events,
and say then therefore with humor that
we're
following Hy's Law.
For (a) and (b) I think the
risk actually,
unfortunately, is potentially high, and
for (c), I
actually think the risk for short-term
use is low,
but has the problem of "hype's"
law, meaning that I
do believe that the creep, the duration
creep that
Steve mentioned is a real issue. There's been so
much hunger for a Coumadin replacement,
as all of
our public advocates have mentioned, that
I can't
believe that this medication would be
used for only
the expected duration. The anticipation of it
being longer should be expected by all of
us due to
the hype and hunger for a
replacement. High, high,
low.
DR. BORER: You've stung me to the quick
by saying that you're at the end of the
line, so
you can start No. 2, which is: Based on currently
available data, is it possible to
identify patients
who are at risk for developing severe
liver
334
toxicity after exposure to ximelagatran?
DR. HIRSCH: Short of some new test, you
know, me profiling things that you've
mentioned,
Mr. Chairman, I think that the greatest
predictor
was exposure to the drug, was in fact
treatment
with the data we were shown. Beyond that, and I
think if I'm integrating the data
correctly, we
would anticipate prior liver disease of
ALT
elevation at the beginning, possibly
statin use,
small body mass index, and females to be
at a
somewhat greater risk, although the truth
is we
need a greater post-marketing exposure to
know if
these are really true.
DR. BORER: Tom?
DR. FLEMING: Just to refer to two of the
sponsor slides, in CS-27 they indicated
if you look
at ALTs three times upper limit of normal
where
treatment has a relative risk of 6.8,
none of the
other prognostic factors had a relative
risk of
even 2.
So it's difficult to really nail down who
this population is. And in their slide CS-30 they
said there is no patient subgroup
identified at
335
higher risk of developing severe liver
toxicity.
DR. BORER: Dr. Sjogren?
DR. SJOGREN: Well, when you confront a
group of patients you really don't know
who's going
to have the idiosyncratic response to the
drug, but
I would think, like with other
medications that we
use in liver disease like Imuran(?) or
others, we
do testing. We initiate the drug, and now with
Imuran we do the testing before because
we have
some markers now, but in the past we did
very
rigorous testing the first month, and we
were able
to then discern what patients we needed
to take the
drug away. So I would think that we now don't
know, but once you start the drug you
can, not
monthly, but probably weekly, monitor
those
patients for a period of time and then
decide who
were at risk of exposure to the drug.
DR. BORER: I think that counts as a
possibly, a possibly yes.
DR. SJOGREN: Yes, possibly yes.
DR. BORER: John?
DR. TEERLINK: I think I'll have to say
336
no, and I'll just leave it at that since
we've
already said.
DR. BORER: Yes.
I'm not a hepatologist
but my response is really along the lines
of what
Dr. Sjogren says. I don't want to obviate the
possibility or exclude the possibility
that there
is a monitoring algorithm that could be
used. It
may be extraordinarily conservative. It may be
difficult to apply. I don't know.
And I certainly
don't know how you would pick out those
patients
right now. I just don't have the knowledge base to
suggest how you could do that, but it
seems to me
that it might be possible to develop an
algorithm
that would allow you to at least reduce
the risk
below what we think it is now in a
population that
might receive the drug, and I would urge
the
sponsor and the FDA to work on developing
such an
algorithm.
Paul, do you have any thoughts
about that?
DR. WATKINS: Not about the algorithm, but
for none of those drugs that I've listed
is there a
way to identify what patient is
susceptible to the
337
severe liver injury, and that's one of
the reasons
the NIH has established this drug induced
liver
injury network to begin collecting
well-defined
cases, make a patient registry and get a
DNA bank.
In the case of ximelagatran, it
looks like
there's a clear susceptibility window
which
suggests to me a genetic component, and I
would
applaud any efforts the company would do
either
retrospectively or prospectively to get
genomic DNA
to help us get some of the answers to
these
questions.
DR. BORER: Steve?
MR. NISSEN: I appreciated the explanation
of the origins of idiosyncratic, and
clearly it
does in fact apply here, and that's the
nature of
idiosyncratic drug reaction, is that it
is not
predictable, and, boy, I sure wish it
were because
it would turn this whole thing around if
somebody
could come up with a way to identify. It's a very
small number of people to have this
problem, but we
just can't pick them out.
DR. BORER: Let me, if I may, just modify
338
what I said a little bit. This question, as it's
stated literally says: Is it possible to identify
patients who are at risk for developing
severe
liver toxicity after exposure? And so I guess you
have to be a little more precise and tell
us
whether that identification has to occur
before any
drug is given, or after some drug is
given. If
it's before any drug is given, we've
heard that
there's no way to do it. If it's after perhaps
some drug is given, maybe there's a way
to identify
people who are responding. Maybe there is, maybe
there isn't.
MR. NISSEN: Jeff, that's sort of in the
next question.
DR. FLEMING: Oh, okay, sorry. Then let's
go on to Susanna.
DR. CUNNINGHAM: I'm going to say no, but
I'm also going to use this as an
opportunity to say
that there's little or no, really not
much data on
ethnicity, diversity in this group that's
between
88 and somewhere in the mid 90s,
caucasian. And so
if there's any further research done on
this, I'd
339
like to encourage the sponsor to look at
some
ethnically diverse groups, especially
African-Americans.
DR. BORER: You know, to point out and get
it on the record here, as Susanna's
point's very
well taken because if you look at the
data we have,
it sounds as if there is an ethnic
difference in
response, and that the Asian population
did not
seem to respond as--with the same
likelihood of
toxicity as other populations. Whether that's an
artifact of sub-analysis or not, I don't
know, but
it's an important point.
Jonathan?
DR. SACKNER-BERNSTEIN: I would say no,
although it's also important that
patients who
started out with an ALT above two times
normal I
think in the studies were excluded, so
that would
be important to include.
DR. BORER: Beverly?
DR. LORELL: I would say it's not possible
prior to initiating drug, with that
important
exception, to identify patients at risk.
340
DR. BORER: Bill?
DR. HIATT: Just add another note of that,
but if you look at the risk factors that
were
presented, those relative risk increases
were
really modest, so I'm thinking about
positive and
negative predictive value around those
things which
would probably leave a lot of margin for
error. So
I think those risk factors just aren't
strong
enough except drug itself to predict a
population
at risk.
DR. BORER: Ron?
DR. PORTMAN: Well, if it's truly
idiosyncratic, then what I'm about to say
is
incorrect, but as we saw earlier, I mean
this is
not a fixed dose for patients who have
chronic
kidney disease, and we saw that as
patients' GFR
dropped that the prevalence of hepatic
enzyme
elevation at least went up, and so that's
certainly
for, as the company admits, for those who
are on
dialysis it's a risk, and so I think that
those
patients who have chronic kidney
disease--and there
are now five stages of chronic kidney
disease that
341
the
company may want to look at for dosing
recommendations, that those may be at
risk for the
higher grades of chronic kidney disease.
DR. BORER: Tom?
DR. PICKERING: No, I don't think any of
the predictors were strong enough to be
of much
practical value.
DR. BORER: Any comment, Dr. Vega?
DR. VEGA: No. I
agree on what Dr.
Watkins said, it's not unique to
ximelagatran; it
is basically for all potentially liver
toxic
molecules.
DR. BORER: Having satisfied Alan, we'll
now go on to our old way of doing things
with
Question 3. Did the sponsor study procedures for
monitoring and managing patients with regard
to
liver function adequately minimize the
risk of
severe liver injury and liver failure in
the
clinical studies?
We'll start out with Steve as
the
Committee reviewer, and if anyone has
anything to
add to his comments, you can. If you just want to
342
vote on this one, which is probably
relatively
straightforward, you can do that too.
Steve?
MR. NISSEN: Yes and no.
I mean let me
see if I can specify what I mean. I was actually a
little surprised that only 70 percent of
the
monitoring that was mandated in the trial
was
actually done, and I say that because when--I
mean
I do a lot of clinical trials myself, you
know, and
those coordinators of ours, you know,
they get
after the patients if they don't show up
for an
appointment to have a blood draw. And so this was,
in the clinical trial, I would have
thought a very
optimal setting in which to get a very
high rate of
compliance with monitoring, and it was
only about
69 or 70 percent, and so it does--to me
it's a very
important question that the FDA is asking
here,
because if the best that we can do in a
controlled
clinical trial is 70 percent, then the
question is
what's going to happen in general use?
And the lowest figure I heard
was the
troglitazone figure of 5 percent, and so
I'm
343
surprised it wasn't 90 percent in the
clinical
trial.
So that part I don't really understand it
based upon my own clinical trial experience. But
having said that, I think the 70 percent
is sort of
intermediate. And so the monitoring procedures
were reasonable, but they were certainly
not as
high as I would have liked to have seen,
and
probably are going to be a lot lower if
it were to
be in general clinical use.
DR. BORER: So did they adequately
minimize the risk?
DR. NISSEN: I mean it's--you know, no. I
mean I think that adequately minimizing
the risk
would have involved much more intensive
and a
higher compliance rate. I just don't understand
why that wasn't achieved.
DR. BORER: Tom?
DR. PICKERING: I agree, no.
DR. PORTMAN: No.
DR. HIATT: No.
DR. LORELL: I think no.
I think it was
very appropriate that the sponsor
modified their
344
algorithm for monitoring, but I will
agree I was
somewhat surprised at the compliance
issue.
DR. SACKNER-BERNSTEIN: No.
DR. CUNNINGHAM: No.
DR. BORER: No.
DR. TEERLINK: No.
DR. SJOGREN: I think the protocol was
good, but the implementation was not
done. So I
guess it's no.
DR. BORER: Tom?
DR. FLEMING: No.
DR. HIRSCH: No, but we're replacing now
ideally the challenge of monitoring
Coumadin with
warfarin clinics with the challenge of
following
ximelagatran with liver health
clinics. It's
better to be liver than deader.
[Laughter.]
DR. BORER: I was going to start on your
end, but I won't now.
[Laughter.]
DR. HIRSCH: It's my last meeting.
DR. BORER: Okay.
Do you have any other
345
safety concerns regarding the long-term use of
ximelagatran, for example, cardiac, and
perhaps
regarding a short-term use of
ximelagatran?
Tom, why don't we start with
you?
DR. FLEMING: There are two other domains
here that I think for me are
noteworthy. One is
the domain of coronary artery disease and
specifically MI, and I'm speaking now
first in the
long-term use. The slide 0-49 which extends the
slide, Table No. 18, that Jeff had called
to the
attention of all of us, Slide 0-49 the
sponsor put
up even extended a bit the completeness
of the
follow up and the MI data that we have in
the 6,900
versus 6,200, and there is an excess of
half a
percent in MIs.
So there is in fact certainly a
real
indication that there is a relationship
here. It
was interesting, when we look at the
fatal MIs in
the SPORTIF III and V trials, in SPORTIF
III it was
11/12, but in SPORTIF V it was 10/3, so
there were
three times as many fatal MIs in the
SPORTIF V
trial.
So it does in fact, to my way of thinking,
346
look consistent across series of data
that there's
something here that's real.
In the bleeding domain, whereas
the
sponsor presented results in the SPORTIF
trials
indicating lesser--somewhat less,
non-significantly
less major bleeding, there were still
twice as many
in this long-term database who
discontinued due to
bleeding, 83 versus 43, again, another
half a
percent.
In the short-term trial as well, there
are, as we have seen, indications of an
excess of
MIs, 16 versus 4, and major bleeds, 18
versus 10,
again consistent with what we're seeing
in the
longer term.
So I would add those two
domains as areas
of concern beyond the hepatic toxicity.
DR. BORER:
You're forgiven, Alan, you can
go ahead.
DR. HIRSCH: Thank you for the
forgiveness.
No, I have the same concerns,
but I do
want to make a position statement. We've been so
worried about adverse effects. It's clearly a
347
positive signal of benefit that's
consistent
through all the trials for preventing
thrombotic
events in two extremely vital
circulations, the
brain and the lungs. Though we have this relative
signal of cardiac thrombotic events that
concerns
me, I believe we might need more data to
show net
benefit, or we actually probably have
data showing
net benefit regarding the other
thrombotic areas.
DR. SJOGREN: I do have some concerns
about MIs. You know, I heard you discuss about it
and I am concerned somewhat about it.
DR. BORER: John?
DR. TEERLINK: I also share the
cardiovascular concern, and it's a low to
moderate
level concern because I think it can be
addressed
in further studies and looked into
further, and may
be able to be eliminated by some additional
data,
but for now I think it's an unaddressed
issue.
DR. BORER: This is for both short- and
long-term, right? I have a couple of concerns but
they don't rise to the level of
showstopper type
concerns, and I have to explain why I say
that.
348
The issue of cardiovascular events is a
concern if
it's real. And I agree with John. I think that
probably the importance of this can be
resolved
with some additional data, but we have
right now is
what we have. On the one hand we have some
observations suggesting an excess of
cardiovascular
events, both MIs, small number, low
absolute risk
and some other generally softer
cardiovascular
events, and heart failure. These events seem to be
excessive in populations in whom they
were
unexpected, and if they were unexpected,
then
finding them is less compelling to me
than if I had
expected them.
But in the situation in which
these kinds
of events specifically were being looked
for as a
matter of protocol as efficacy issues, I
didn't see
that excess. In fact, there was a tendency for
things to look better on ximelagatran
than on the
comparitors. So where I expected ximelagatran to
look good, it seemed to look good. Where I didn't
expect it to look bad, it maybe looked
bad. I
don't know how to put all that together.
349
In any event, the absolute
number of
excess events seemed to be small, which
is why I
say I'm not overwhelmingly concerned, but
concerned
enough to look at little further. There are
several confounders that may be involved
here. One
is the possibly differing pathophysiology
of the
coagulation system and the cardiovascular
system in
the different populations, early post-op
patients,
et cetera, et cetera, versus non-operated
patients.
We've raised the concern about rebound
which I
think has been discussed enough, but it
still is
sitting there as something I'm thinking
about.
And then, of course, we have to
determine
if there really is some excess of events
on
ximelagatran, whether that represents a
toxic
effect or a lack of benefit because of,
as Steve
pointed out, a differing effect on venous
versus
arterial thrombotic processes. That seems a little
farfetched, but it could be.
So I have some concerns, but
they don't
rise to the level of showstoppers. With regard to
the bleeding events, it seemed to me that
the more
350
drug you give in general of an
anti-thrombotic, the
more likely you are to bleed, and overall
these
data look that way to me, so that doesn't
concern
me too much. So I do have some concern, but I
would think about it in the context that
I just
presented it.
Steve?
DR. NISSEN: I guess we're going to skip
you, Paul, or maybe we'll come back to
you.
First of all, I want to
reassure my
friend, Tom Fleming, about the bleeding
issue, and
let me tell you why I'm a little bit
reassured
about it.
That is that I agree with the sponsor
and their representatives that the use of
Coumadin
in this trial was probably more precise
and better
controlled then we see in the general
use. So the
fact that Coumadin did pretty well on
bleeding is
not what we actually see in the real
world. In the
real world it's a lot sloppier. And so there would
likely be some equalization here on the
bleeding
rates when you consider how drugs are
generally
used, particularly Coumadin, are used in
real life
351
where the monitoring just isn't as good
as we'd
like it to be.
However, I have considerable
concerns,
more than others, about the cardiac, and
let me
tell you why. Yes, the absolute numbers are low,
but in the short-term trial the exposure
duration
is very short, and so if I look at an
event rate of
the magnitude that we saw for myocardial
infarction
with 12 days of exposure, it starts to
look a lot
worse, and it maybe looks even a little
bit scary
there as you sort of think about it. It is
statistically significant.
Now, p-values don't mean that
everything
is biologically significant, but it sure
gives you
the bias that it probably is. So when I see a
statistically significant excess of
cardiac events,
and when the magnitude of the events is
high for
the duration of drug exposure, it makes
me worry.
And I think it would need to be explored
before it
would be safe to use that drug in a
population of
old people who are undergoing knee
replacement and
hip replacement, have a lot of
concomitant
352
cardiovascular risk factors, and I know
what these
people look like because I look at
them. A lot of
them are going to have coronary
disease. And so I
think we just don't know, but it doesn't
look
promising in terms of the cardiovascular
risk.
DR. BORER: Paul, I didn't mean to skip
over you.
If you have any comment? No?
Susanna?
DR. CUNNINGHAM: I agree with Steve.
DR. BORER: Jonathan?
DR. SACKNER-BERNSTEIN: I agree with Steve
about the risks, and I would say that I
feel
reasonably similar to him about the
cardiac risk.
I think that it's fair to pull in a paper
that Bob
Temple wrote a few years back, where he
was quoted
in that paper saying that a signal with a
p-value
of .1, potentially showing a safety risk
when
you're looking at a database even post
hoc, is a
signal that's worth enough of a concern
that it
nearly needs to be addressed
formally. So I think
that this level of a signal, this kind of
p-value,
means it's something that really deserves
some
353
future attention.
I would also say that the
long-term a-fib
population is a concern in two
respects. One, I
agree that the manner of management of
Coumadin was
much better than seen in clinical practice,
but I
would also expect that the management of
liver risk
is better than will be seen in clinical
practice.
So all we can do is really compare apples
to apples
and say in this trial, this is how the
warfarin was
managed, this was how the liver was
managed. Let's
just balance the risks between the two.
And that also becomes important
when you
start to look at the kinds of patients
seen in
clinical practice who have atrial
fibrillation. I
don't know the demographics perfectly,
but I know
that there's a huge proportion of a-fib
patients
over 75, and I know this data a little
bit. But if
you look at a typical patient who's 75 or
80-years-old, unless they have a
creatinine down at
about .7 or .8, they're going to have a
creatinine
clearance that's very, very different
than the
creatinine clearances of the patients in
these
354
trials, and you're going to see patients
treated
with this drug that have calculated
creatinine
clearances of 30 to 45 cc's per
minute. That's an
area where we don't know what the
risk/benefit
ratio is and there really aren't
sufficient data.
I think it's an area that if you're going
to be
advocating this drug be considered for
atrial
fibrillation, really have to do a primary
analysis
in people who have impaired renal
function.
DR. BORER: Beverly.
DR. LORELL: Thank you, Jeff. My feelings
about the cardiovascular risk in
long-term use are
more similar to yours than to
Steve's. I think
that that very small apparent absolute
increase is
in somewhat a hypothesis generating. I don't see
it as being an extremely strong or hard
signal that
there is an increased risk.
My thoughts about bleeding are
similar to
yours, Steve. I agree with you on that.
To my mind, the short-term
signal I think
may be a real issue. As I questioned earlier
today, I think it relates to a more
generic problem
355
that all of us around the table have seen,
interventional cardiologists certainly
see. We
know that the post-surgical period for
any kind of
lower-extremity or hip surgery is a
period of
prolonged tissue damage and inflammation,
and we
know enough in 2004 to realize that is a
milieu
where there is a higher risk of
vulnerable crack
rupture, of acute coronary syndrome and
infarction.
So to me that signal of a possible,
quote,
"rebound," I don't look at so
much as rebound, but
being the more generic issue of how you
manage
aspirin correctly and fastidiously in
post-surgical
patients who have risk factors for
coronary disease
period.
And I think that's a more generic issue.
So that's a little bit of a
long-winded
answer, but I did want to make the
comment that I
look at it somewhat different as being a
narrow
rebound issue as opposed to a broader
issue of the
importance of using aspirin carefully in
this
population.
DR. BORER: Bill?
DR. HIATT: I think there is a predictable
356
bleeding risk, and I'm not really concerned
about
it because it matches the pharmacology of
the drug,
it's related to levels in the
circulation. So I'm
really not terribly concerned about the
bleeding
risk.
I agree with everybody else
around the
table, there may be a cardiovascular
risk, and I
have two questions about that. One is:
were those
events adjudicated, the MIs? I don't remember if
we asked that question.
DR. SHETH: The events in the SPORTIF
trials, SPORTIF III and V, and the events
in the
ESTEEM trial were pre-specified and
adjudicated,
and the SPORTIF trials did not
demonstrate a
difference with that comparitor, and you
know the
results of the ESTEEM.
DR. HIATT: Yes.
DR. TEERLINK: Thanks for clarifying that.
I think that was a point that if we
really want to
believe that there's a cardiovascular
signal, I
would really want to have some hard data,
so I'm
not yet fully convinced that there is.
357
I think the other challenge
that we
brought up earlier, if there is such an
increased
risk, to try to quantify that might be a
difficult
thing to do in a population for whom
you're not
expecting a lot of events. And what I'd ask is not
an assessment of the mean event rate
difference
between drug and placebo, but the 95
percent
competence interval, the upper end of that
competence interval around that
risk. And where
would you set that? So I think to say there might
be something there leads you to the next
question,
how would you then go quantify that, and
what level
of risk would you be willing to accept if
in fact
it's there?
DR. BORER: Ron?
DR. PORTMAN: I think the bleeding is an
issue, particularly in the advanced CKD
population,
but--and I also would like an answer to my
question
related to the CKD stage and the
cardiovascular
complications. I am concerned about that, but I
would agree with you that it's not really
a
showstopper. I think it's an opportunity. If we
358
can figure out the mechanism of why this
seems to
be happening, it would be really
fascinating, and
perhaps come up with ways to manage that.
DR. BORER: Do you want an answer now to
the question you raised?
DR. PORTMAN: If she has it.
DR. SHETH: Yes.
We did manage to--
DR. BORER: Just a quickie.
DR. SHETH: Okay.
0-50, please. I'm
sorry, I don't have my screen in front of
me. Here
we go.
Calculated creatinine clearance on the
right-hand side, normal renal function,
mild renal
impairment, moderate and severe, and
whether or not
patients had an MI, yes or no. Ximelagatran,
warfarin, placebo.
And you see that for patients
without
renal impairment the incidence is about
40 percent
across, 44, 37 and 40. For patients in the mild
renal impairment group it's 39 for
ximelagatran,
37.5 for warfarin and 43.8 for placebo.
Then if we take a look at
moderate, it's
11 percent for ximelagatran, 25 percent
for
359
warfarin and 15 percent for placebo. And in the
severe group the numbers are too small.
DR. BORER: What are those numbers in
parentheses?
DR. SHETH: The numbers in parentheses, I
believe, are the percent.
DR. BORER: Percent of what?
DR. SHETH: But what I don't have--I'm
sorry.
Just understand that this is not the
percent over the total population. This is the
percent of patients who had normal renal
function
and had an MI, so the denominator is not
the 6931.
It's the percent of the column
total. Sorry.
Thank you.
DR. BORER: Okay.
Ron?
DR. PORTMAN: That's not what I would have
expected, but that's interesting. Thank you.
DR. BORER: Tom?
DR. PICKERING: One thing about the
bleeding of concern is the absence of any
antidote.
You can't give these patients vitamin K
if they
come in with a massive bleed. I guess you just
360
have to wait and hope.
With regard to the MI, I'm not
that
concerned. The numbers were small. It was in
EXULT A but not really in EXULT B, and I
was
reassured by the ESTEEM and SPORTIF
data. So I
think there may be something there, but
it's not
something that would really have much
influence on
my overall decision.
DR. BORER: Dr. Vega?
[No response.]
DR. BORER: We've now finished Section I.
We have not all that much time, so I'll
ask to give
shorter answers, and I am the most
egregious
non-doer of that, so I will shorten my
answers too.
Short-term use. Now we're talking about
benefit risk, specifically short-term
use,
prevention of VTE in patients undergoing
elective
total knee replacement surgery.
No. 1. Do you recommend additional safety
studies with longer follow-up to address
the
possibility of delayed occurrence of
liver toxicity
following short-term use?
361
Steve?
DR. NISSEN: Briefly, I'd like to see
exposure up to, say, 30 days, with
another three
months of follow-up because I think that
that kind
of duration creep is likely to occur in
clinical
practice, and so we would need to know
exactly what
happens if you get 30 days of exposure
over the
next, say, three months thereafter.
DR. BORER: Okay.
Alan?
DR. HIRSCH: Agreed.
DR. BORER: Tom?
DR. FLEMING: Agreed.
DR. SJOGREN: Agreed.
DR. :
Agreed.
DR. BORER: Me too.
DR. CUNNINGHAM: Me too.
DR. SACKNER-BERNSTEIN: Agreed, but I
would urge that if the comparitor is
warfarin, that
actually the starting doses are according
to the
ACCP guidelines of perioperative
treatment,
starting at 5 to 10 milligrams instead of
2-1/2 as
were done in this study.
362
DR. LORELL: I agree.
DR. HIATT: I do too, and if I had that
information, that would resolve my major
concern
about approvability for that indication.
DR. BORER: Ron?
DR. PORTMAN: I agree.
DR. PICKERING: Yeah.
DR. BORER: No. 6.
Regarding the
potential risk of myocardial infarction/coronary
artery disease with short-term exposure
ximelagatran (mean 8 days) in patients
undergoing
total knee replacement, do you recommend
further
studies to assess the risk of acute
MI/CAD? If
yes, what type of studies do you
recommend?
I think Bill discussed this
already. Do
you have sufficient information from our
comments
about that question, or do you want us to
go
through that again?
DR. KORVICK: I think it might be helpful
to vote, and then if there's a specific
design that
you wanted to vote on, to just make it
crisp,
because the discussion was long. I think just
363
crisp it up.
DR. BORER: Okay.
Steve, can you crisp it
up?
DR. NISSEN: Very crisply.
Again, I'll
suggest a design, which is because
there's a
statistically significant p-value in this
short-term population, I can't make this
go away
without additional data. Therefore, I would
suggest studying patients that are at
high risk,
patients that have had a prior MI or have
known
vascular disease, that is, arterial
vascular
disease, and try to confirm or refute,
and
adjudicate the events carefully. That doesn't have
to be a huge population because the
higher risk
group that you study is going to have a
higher
event rate and therefore is going to
accrue more
events.
But I think you can target a high-risk
population and answer the question in
this surgical
group.
DR. BORER: Does anyone have any other
opinions besides what we heard? Jonathan and then
Beverly.
364
DR. SACKNER-BERNSTEIN: As I said before,
I think that it is possible that the risk
of
infarction is related to the fact that
these
patients are untreated, and as Beverly
said before,
have a pro-inflammatory state in
surgery. I would
argue that the trial should be one in
patients who
appear to be low risk, who are not on
cardiovascular medicines, because that's
the only
way you're going to address the signal
that we see
so far.
DR. BORER: Beverly?
DR. LORELL: I disagree.
I don't feel
that there is an indication for another
study with
short-term exposure. I do feel there is a need in
education after approval and in labeling
to remind
physicians to use aspirin appropriately,
as current
guidelines indicate.
DR. BORER: Any other opinions? Alan?
DR. HIRSCH: I'll try to be creative. I
believe like Steve the signal requires
some
additional data, but it is very
challenging to
design and offer a design that's
practical. This
365
is a tantalizing molecule that looks like
it should
have therapeutic use. I'd like to define the
population which is clearly safe. The average
25-year-old or the 35-year-old at low
risk, that
has a very low risk of MI, would be
tantalizing to
use this medication, and so I would do a
pre hoc
risk assessment before the orthopedic
procedure and
define risk by pre *hoc parameters,
Framingham
parameters, whatever you'd like.
DR. BORER: Yes, Norman?
DR. STOCKBRIDGE: Quick question for
Steve.
Do you have any idea what dose you'd want
to use in this new trial?
DR. NISSEN: I assume we would just test
the 36-milligram dose because isn't that
the dose
that's being requested here?
DR. STOCKBRIDGE: You have some sense that
that was a sensible thing for them to do?
DR. NISSEN: Yes. I
mean I don't--maybe I
missed something.
DR. STOCKBRIDGE: Are you in fact trying
to confirm a phenomenon or are you trying
to figure
366
out how to minimize the effect?
DR. NISSEN: Oh, I see.
Well, first of
all--
DR. STOCKBRIDGE: Haven't had a maximized
benefit.
DR. NISSEN: I understand what you're
saying.
I see a signal which is a statistically
significant excess of events that may be
a false
signal, and therefore, I am looking for a
study
that will confirm or refute that the
doses used in
the short-term trial will result in
excess
short-term cardiovascular risk.
DR. STOCKBRIDGE: So if you get a yes/no
answer, you're still not going to know
what to do
about it without studying dose.
DR. NISSEN: You will not know what to do
about it, but I guess I think the risk is
high
enough that we're getting a false signal
here, that
there really isn't any increased risk;
that I am
looking for a confirmatory study that
would
indicate that there is, because if we
could make
that excess risk go away, it would
increase my
367
comfort level with the approvability for
the
short-term indication.
DR. BORER: I'll add just one point here,
and that is that I don't know that it's
an
appropriate use of resources to do
another study in
people for 8 days. I think that Steve's earlier
point is the important one. We're concerned really
not about 8 days but about the protracted
use that
will probably be the model in clinical
practice.
So I would want to see the study that you
suggested
earlier, which I think would provide
information
about whether these unexpected results
are repeated
in a post operative population or not,
and I'd
probably use the dose that the sponsor
has
suggested is effective.
There are questions that can't
be answered
if we use that design, but my intuition
is that
that will provide us with the most
information and
the key information that we need to be
able to know
about the most important risk which is
liver
toxicity if the drug is used in a more
prolonged
regimen.
368
DR. HOUN: And that 30-day study, you're
saying you're recommending it
pre-approval?
DR. NISSEN: I am.
I would not be
comfortable without additional data.
DR. HIATT: Could I comment on that? I'm
sorry.
Go ahead.
DR. AVIGAN: I was just going to ask
another study design question, which is,
remember
the comparitor was warfarin in the 8-day
study.
Would you, in the safety study, as for a
basis of
comparison want to look again at warfarin
for the
longer duration, or what would be the
comparitor,
and how would you construct the trial?
DR. BORER: I'll take a stab at that, and
then people can disagree. I think that warfarin is
a reasonable comparitor because it's used
commonly
clinically. I don't believe that the statements we
heard or read about the comparison being
unfair are
really germane. This is a recommended treatment by
a consensus panel, even though it's not
an approved
indication. It is used clinically. Such a study
would provide us with a great deal of
information,
369
and the fact that warfarin doesn't get
you to your
therapeutic level so fast is a
problem. It's a
problem in the use of warfarin. So in fact, that
comparison provides us with useful
information. If
the sponsor chose to use low-molecular
weight
heparin followed by warfarin, that would
be fine
too, but I wouldn't find it unacceptable
to look at
warfarin alone.
Bill?
DR. HIATT: This is analogous to the
discussion on cilostazol several years
ago, where a
drug is being approved for symptomatic
indications
that may or may not be a signal of excess
mortality, but the event rates were very
low.
I think what has to happen here
is you
have to look at this putative MI risk,
look at the
confidence interval around the difference
between
drug and warfarin, and then look at
the--which has
got to be large now, and then ask the
question:
how
many patients do you need to shrink that
confidence interval so the upper limit
end of the
confidence interval is below some
threshold. What
370
risk are you willing to accept if it's
real, 25
percent, 50 percent? I mean I think you have to
get rather specific about how to address
this
question, because then it would be simply
a number
of patients needed to study to look at
the
confidence interval around any putative
risk in
this particular area. So I would just play with
those numbers and set that up.
Now, the question is, do you
want to do
that before or after approval? It depends on your
experience in that area.
DR. BORER: Which brings us really to the
next question. I'm sorry?
Okay. I thought I
could get away with that.
[Laughter.]
DR. BORER: Okay.
Tom, yes or no on
Number 6?
Do you recommend further studies to
assess risk of acute MI/CAD? If yes, what?
DR. PICKERING: No.
DR. BORER: Ron?
DR. PORTMAN: Yes, the same as in No. 5.
DR. BORER: Bill?
371
DR. HIATT: Yes.
DR. BORER: Beverly?
Actually, you
already gave your opinion. And so did you, Jon.
DR. LORELL: I do, however, advocate doing
the study for roughly 30 days to
understand the
liver issue.
DR. BORER: Jonathan, you already voted.
Susanna?
Oh, you did? Sorry.
DR. CUNNINGHAM: No. It
would be yes if I
did.
DR. BORER: It would be yes. John?
DR. TEERLINK: Yes, with moderate risk
cardiovascular patients followed for at
least 30
days of treatment and followed for three
months at
least.
DR. SJOGREN: No.
DR. FLEMING: Yes.
DR. HIRSCH: Yes.
DR. BORER: I said yes, but the 30-day
study.
Based on the currently
available data, do
you conclude that the benefits of
ximelagatran for
372
short-term use for prevention of VTE in
patients
undergoing elective total knee
replacement surgery
outweigh its risk? So this is the sort of
approvability question, and the question
about
whether the study we're all talking about
needs to
be done pre-approval or post-approval can
be
inferred from the answer to this.
Why don't we start at that end
of the
table?
Tom?
DR. PICKERING: I would say no to this one
on the grounds that it was marginally
better than
warfarin for the venography endpoints,
but I think
the optimal standard of care would be low
molecular
weight heparin which has not been
compared against,
and there was no suggestion that the
major events
like PE and MIs were reduced.
DR. BORER: Ron?
DR. PORTMAN: Based on my answers on 5 and
6, I would have to say no.
DR. BORER: Bill?
DR. HIATT: Technically no at this point,
but I think I said earlier, if that
safety data
373
were available, I think they proved their
primary
endpoint, the validity of that, but it's
a positive
study.
So if the safety issue can be resolved,
then I would vote for approval.
DR. BORER: Beverly?
DR. LORELL: That's precisely my answer as
well.
DR. BORER: Jonathan?
DR. SACKNER-BERNSTEIN: I would also vote
no because the risk really hasn't been
defined
adequately to establish a risk/benefit
ratio.
DR. BORER: Susanna?
DR. CUNNINGHAM: I agree, and I don't
think the benefit's large enough.
DR. NISSEN: I'm sorry, but I'm answering
a little bit longer here.
DR. BORER: You're allowed. You're the
Committee reviewer.
DR. NISSEN: I know.
This is a little
tougher question. I want to make sure I get on the
record why I'm voting the way I am. First of all,
in terms of benefit what we have actually
is excess
374
risk for the heart events. If you take the
composite of the things that you really
care about,
death, pulmonary embolus and MI, there
are more
events in the ximelagatran arm than the
comparitor
arms in the short-term studies. So the proof of
benefit for the things that really are
the most
important is simply not there, it goes
the wrong
way.
Secondly, we really don't know
what
happens if you give this drug for longer,
and I
have to believe that it's going to be
given for at
least up to 30 days after an event like
this. So
now we have the added problem of not
really knowing
what the risk is of a 30-day exposure
over the next
three months.
Third point. Warfarin is not burdensome
for 30-day administration. We've got a drug out
there--I mean I heard all the arguments
about the
guy in Montana, but, you know, they're
much more
applicable to somebody who's on it for
years than
somebody who's on it for 30 days. It's not a big
deal to take warfarin for 30 days, and so
I don't
375
think we have a deficit in clinically
available
alternatives, and that's why we need to
be more
secure in our information about what's
going to
happen when this agent gets out there.
I'm very worried about the
duration creep,
and I'm very worried that if we let the
genie out
of the bottle for this very limited
indication, it
would be very hard to put it back in
again. So
before I'm willing to let the genie out
of the
bottle, I want to make sure I know what
the liver
risk is in a 30-day exposure, and without
knowing
that, I can't be comfortable with the
agent.
DR. BORER: I agree with Steve. It's
important for me to say that my intuition
is that
this drug is useful for short-term use
and probably
for long-term use if we could deal with
the
toxicity issue a little bit better, but I
am
concerned that I'm basing that intuitive
judgment
on my belief about pathophysiology and my
inferences about what the natural history
would be
on that basis rather than on a consistent
result
with heart events and pseudo surrogates.
376
So I'm a little concerned about
that. I
would like to know more about the safety
with the
30-day business, and therefore I
don't--although I
think ultimately this drug should be
approvable for
the indication, I don't think we have
sufficient
information right at this moment to allow
that to
happen.
John?
DR. TEERLINK: Jeff, you skipped Paul
again.
DR. BORER: I'm sorry.
Paul?
DR. WATKINS: No, no, I have nothing to
say.
DR. TEERLINK: I share your enthusiasm for
the medicine, and hope that it will
someday be able
to be used for many of these indications,
although
now I can't say that it would be
appropriate for
this indication for two reasons: because I'm not
convinced of the endpoint in terms of how
it played
out with being a surrogate, and secondly,
the risk
I don't think it well enough defined at
this time.
DR. SJOGREN: I agree.
My answer is no.
377
DR. FLEMING: I would like to give some
very brief specifics to justify my
answer. I think
Jonathan Halperin's discussion of saying
in the
Indication 2 that we should focus on what
does the
data show in the aggregate on the primary
endpoints, major bleeds and death. He didn't
actually do that, although I did for this
knee
replacement surgery indication. If you start at
deaths it's 10/4 in the wrong
direction. If you're
adding PEs and MIs now you're capturing
what I
would consider to be the most significant
major
events; it's 31/14 in the wrong
direction. If we
add, as he did, major bleeds, 2 of which
were
fatal, it's 49/24 in the wrong
direction. We're
still at a doubling of the rate. If we move on and
add symptomatic DVT, it's still 68/49 in
the wrong
direction with a 40 percent excess. If we move on,
guided by Tom Pickering's summary of the
stated
guidelines that some would at least
include
asymptomatic proximal DVT and we add
those in, it's
still in the wrong direction. It's 111/105.
What's left is the asymptomatic distal
DVT, and in
378
my view that's at best a non-validated
surrogate.
So as we look through what are
really the
most significant events, they're in the
wrong
direction, and even including everything
other than
asymptomatic distal DVT, there's still
not even a
positive relationship. So I think guided by Dr.
Halperin's way of looking at this, I
think it is in
fact not established to be favorable
benefit to
risk in this setting.
DR. BORER: Alan?
DR. HIRSCH: It's hard to follow Tom.
I would say no, at the current
time I
don't have adequate data yet to be
confident that
risk is worth the benefit. But I would like to say
I think there is a potential role for
this as a
therapeutic molecule in this indication and
we're
here to opine. I think we haven't probably the
adequate data to know whether this
putative
surrogate marker of thrombosis in the
infrapopliteal segment really is
asymptomatic, and
I suspect, with a well-designed trial,
one might
conceive of an outcome where the signal
of adverse
379
events is less than we anticipate and the
relative
therapeutic symptomatic benefit might be
measurable. So the answer is no now; more data
needed.
DR. BORER: Now we go on to the long-term
use.
Again, I think that we've heard a great deal
here, but for the record we'll go around
the table.
Based on currently available data, do the
benefits
of ximelagatran for secondary prevention
of VTE for
18 months after six months of standard
treatment
for an episode of acute VTE outweigh the
risks for
this indication?
We discussed the risks, but now
we have to
give a specific answer about the
weighting of the
benefits and risks.
Steve?
DR. NISSEN: This gets much tougher for
me,
and let me say what I think hopefully will not
be viewed as excessively harsh, but if
you have an
agent that has serious risks of toxicity
that are
fatal toxicity like liver failure in a
range that's
similar to what we've seen in drugs that
had to be
380
withdrawn from the market, then you're
going to
have to show superiority over existing
therapies.
And so for any of the long-term indications--I've
thought about this very hard in the last
couple of
weeks--I'm convinced that to overcome the
current
burdens of the liver failure problem we
need to see
unequivocal superiority on outcomes of
importance,
that equivalence is not equivalent when
you have a
1 in 2,000 potential risk of fatal liver
injury.
And so I obviously think we're not there
yet.
DR. BORER: Tom?
DR. FLEMING: You're answering Question
No. 8?
So Question No. 8 relating to the long-term
use?
DR. NISSEN: Yes.
DR. FLEMING: You just threw me with your
statement about the equivalence, so I
wasn't quite
following that.
DR. NISSEN: Let me actually clarify that
just--
DR. FLEMING: Which is an issue of No. 10.
DR. NISSEN: Let me just clarify why.
381
Because we have data for 8 which is
superiority
data over placebo, and I answered it the
way I did
deliberately, because basically to make 8
approvable I would want to see
superiority over
another anticoagulant therapy, even for 8
where we
have placebo data. That was why I answered that
way.
DR. FLEMING: This too for me is the
hardest question, and I will articulate
later the
pros and cons as I'm looking at them, but
I'd
surely, in this case, like to hear my
clinical
colleagues' judgments about those first
if you
don't mind coming back to me?
DR. BORER: Alan?
DR. HIRSCH: I'll simply echo more or less
what Steve said. We need an active comparitor and
superiority. The trial design, when it was
created, was appropriate. Currently need a
different comparitor.
DR. SJOGREN: I agree.
I don't think
we're there yet.
DR. BORER: John?
382
DR. TEERLINK: I think the study is very
impressive for demonstrating that we
should
anticoagulate patients long term. Now the question
is
what agent should we use? That's why
it's no
for now, but perhaps yes later.
DR. BORER: I don't think that we have
data sufficient for approval for this
indication
now because the risk is of concern and
it's not
well defined, and we don't know yet
whether we have
an algorithm that might acceptably
minimize the
risk to patients who get the drug. So I'm
concerned about that.
I can't agree that the bar has
to be set
at
superiority in efficacy to Coumadin because
Coumadin has problems too. But before I could
wholeheartedly favor approval, which I
cannot now
for this indication, I would like to see
minimization of risk to a greater extent than
we've
seen, as well as the kind of efficacy
that we've
seen because I think as John just said,
this
represents a practice pattern that isn't
now
generally followed, and these data
suggest that it
383
should be.
Paul, do you have any? No.
Steve?
Steve, you spoke already.
Susanna?
DR. CUNNINGHAM: I don't think I have much
to add except I don't think the public is
ready to
accept liver failure. Livers are hard to come by,
and it's really not a complication they
would be
enthusiastic about.
DR. BORER: Jonathan?
DR. SACKNER-BERNSTEIN: I'm convinced that
it was better than placebo. I'm also convinced
that the study was not designed in
keeping with the
consensus guidelines that exist that
recommend, at
least in the year 2001 when they were
published,
that these patients probably were treated
with
Coumadin for two brief a period of time
before they
were enrolled in the trial, and
therefore, in a
study design they were studying a
scenario that
really is not relevant to best practices. These
patients probably, as best I can tell
from the
guidelines and their characteristics,
should have
384
been treated with Coumadin for at least a
year in
most or at least a lot of cases before being
then
randomized to a new agent or
placebo. So I would
say no, there's not sufficient data.
DR. LORELL: This was also a very tough
one for me. I thought this study was very powerful
in demonstrating the benefit of secondary
prophylactic therapy in this group
compared with
what is certainly the general practice of
not
treating, at least in the United
States. I also
agree that with the issue of the specter
of very
severe and fulminant liver toxicity,
although it is
rare in absolute terms, it's disastrous
if it
occurs to a single patient and their
family. I
would like to see a comparitor study
against
Coumadin, applying the algorithm, the
revised
algorithm as if it were being used in a
larger
population. I'd like to see some evidence that
whatever algorithm is proposed, which
will look
different, I'm sure, than what we've so far,
some
signal that it might be working.
DR. BORER: Bill?
385
DR. HIATT: I would vote no. And I think
the arguments are very well articulated
and I can't
add anything to them.
DR. BORER: Ron?
DR. PORTMAN: I think Coumadin is a
problematic drug, and I think this one
has lots of
potential advantages, but I would like to
see a
study where the company uses whatever
algorithm
they're going to come up with, and see
what it can
do to reduce the liver toxicity.
DR. BORER: Tom?
DR. PICKERING: Just to be different, I
would vote yes. There was substantially less
pulmonary emboli in the patients who were
treated,
and it's all very well to say that these
patients
should be given Coumadin, but in reality
they're
not.
And there was no excess of bleeds.
And I
guess I'm rating the liver toxicity issue
slightly
lower than some of the others.
DR. BORER: Now we go on to the final--I'm
sorry?
Did you want to say something?
Oh, yes,
I'm sorry.
386
DR. FLEMING: Actually, I would have been
comfortable skipping me here on this one,
which is
uncommon for me, but I really have
difficulty with
this.
I try to look at things
quantitatively,
and the major things that I'm seeing are
evidence
per 200 patients of about 6 or 7
prevented
pulmonary embolism as a significant
positive
effect, and against that is, per 200
patients, what
seems to be by our best estimate, one
additional MI
and one additional severe liver
injury. And I have
trouble balancing those pros and those
cons.
I'll abstain.
DR. BORER: That's okay, because we're
going to No. 9 and you're number one.
Is the non-inferiority margin
of 2 percent
compared to warfarin adequate to ensure
that
ximelagatran is non-inferior to warfarin
with
respect to efficacy? If no, what should the
non-inferiority margin be for the
indication of
prevention of stroke and systemic embolic
events in
patients with atrial fibrillation?
387
DR. FLEMING: Well, these are complex
issues, and I'd like to begin by thanking
Lloyd
Fisher, John Lawrence and others for some
very
thoughtful analyses on what is
intrinsically a
really difficult issue to handle in a
most informed
and unbiased way.
So beginning with the first
part of the
answer to Question No. 9, I definitely
concur with
the sponsor that a non-inferiority design
is proper
here in this setting. I think they did the right
trial in the context of a non-inferiority
study
being done.
The 2 percent margin that was
put forward
was, in essence, based on an expected
background
rate of 3.1 percent for warfarin, for its
event
rate, and there was, inappropriately,
when that was
proposed, no adjustment for several
factors. The
first is the constancy assumption,
critical factor.
I'll come back to that. The second is they didn't
adjust for the fact that their estimate
of efficacy
of warfarin itself was variable and not
precisely
known.
They didn't adjust for the need to preserve
388
a fraction of the warfarin effect, such
as half of
the warfarin effect.
And, of course, this part they
didn't know
in advance, that the actual rate wasn't
going to be
in the neighborhood of 3.1 percent, it
was going to
be 1.16 percent. And that's certainly very
significant. If you could justify an absolute
increase of 2 percent and the baseline
rate was 3.1
on warfarin, that would be a relative 67
percent
increase.
That, to me, is pretty substantial.
And
whether that is in fact too liberal a
margin we
could debate, but I don't think it's
debatable that
it's way too liberal a margin when the
background
rate is 1 percent, because now going from
1 to 3 is
a relative 200 percent increase.
Just to give you a sense about why I
think
it's unjustifiable to have used such a
margin, the
actual data they had in SPORTIF V said
there were
excess events, 51 against 37. If there actually
had been 2-1/2 percent versus 1.16,
rather than a
1.61 against 1.16, it had been 2-1/2
against 1.16,
that's 79 events against 37. If these data had
389
shown primary endpoints in excess, 79
against 37,
you still would have satisfied the 2
percent
margin.
Is that clinically relevant? I
struggle
thinking that patients wouldn't care if
warfarin
would have had 37 strokes and Exanta
would have had
79, and we still wouldn't have been
calling that
clinically relevant.
By the way, if we look at it as
a relative
risk, that relative risk would have been
statistically significantly increased
with a Z
statistic of 4, and yet it would have
satisfied
their margin. So the 2 percent margin absolute,
absolutely doesn't in fact make clinical
relevant
sense or statistical sense in the context
of now
applying it in the setting where you're
going to
have a very much lower rate of events in
the active
comparitor arm than what you were
anticipating in
the trial design.
In fact, just to close, in that
example I
gave, the actual data would have been
consistent
with a relative risk on warfarin against
Exanta of
.32, which would have been
bigger--actually, that's
390
consistent with what placebo would have
been, and
yet you would have satisfied that margin.
So what is in fact an
appropriate--the
second part of the question--what would
be a proper
non-inferiority margin? Let me start by answering
this by essentially saying, what are the
factors
that have to guide the choice of the
non-inferiority margin? The first is it has to be
a difference what you can allow that you
would say
is in fact not a clinically important or
clinically
relevant difference, i.e., we're willing
to have a
somewhat lesser efficacy here of Exanta
than
warfarin to a level that patients would
consider to
be an acceptable level of loss of
efficacy. Of
course, what motivates a more flexible
approach
there is when you have a much safer
intervention,
much more convenient to apply or more
cost
effective. Those are the factors that would
influence that decision.
There are also many other
factors. One of
them is this constancy assumption that I
referred
to, and I want to talk a bit more about
that right
391
now.
When you are, as Dr. Fisher pointed out
earlier on, when you're making an
assessment of
efficacy in a non-inferiority trial and
you're
looking at the efficacy against placebo,
there are
really two things that you're looking
at. You're
looking at the efficacy of Exanta against
warfarin
and warfarin against placebo, and the
latter comes
from historical evidence. And they provided six
trials.
The question is: is the estimate of
efficacy of warfarin in those six trials
a reliable
estimate of what the efficacy of warfarin
is in the
Exanta trial? That's the constancy assumption.
Well, why wouldn't it be?
There are several reasons why
it might not
be.
Were the populations the same in those six
trials as in the SPORTIF trial? Was supportive
care the same in those trials? In fact, if
supportive care in a U.S. setting in the
SPORTIF V
trial was more enhanced, you could
readily expect
that the additive effect of warfarin to
that
supportive care could be less.
392
What about issues of how the
outcome was
assessed?
All of these are things that influence
the validity of the constancy
assumption. One
thing that makes me worry a lot about the
validity
is if you look at the statistical
analysis that was
provided by the FDA in Table 1 on page 7,
the
listing of the event rates in the
warfarin arm in
the six historical trials reflect much
higher rates
of events than in the SPORTIF V
trial. It
certainly is a smoking gun for suggesting
that the
validity of the constancy assumption is
at best
uncertain.
There are other issues as well
when you do
a non-inferiority trial. You have to make sure
that the active comparitor was delivered
in a way
that maximized its efficacy, because the
best way
to look equal to something that's an
active
comparitor is to deliver the active comparitor in a
way that's not particularly optimally
delivered. I
suspect that probably isn't a key issue
in this
case.
And then there's the issue of
blinded
393
assessments. What was also apparent in this FDA
review is not only was the SPORTIF V
trial showing
a lesser effect as the blinded trial, but
in the
six warfarin trials, where two were
blinded, those
two also showed lesser effects. So that adds,
obviously, some additional complexity.
A second issue that you have to
consider
when you define them--or a third, I
guess, because
the first was clinical relevance and the
second was
a validity of the constancy assumption,
is you've
got to take into account the uncertainty
or the
variability in the estimate of the active
comparitor's effect, and you have to take
into
account how variable were those estimates
of
efficacy across trials?
Next issue is the issue of
blinding, and
how in fact do we think that influences
our overall
estimates is something we have to take
into
account.
And finally, I think--and it
actually
emerged from this Committee many years
ago--there
is a sense or a standard that if you're
going to
394
replace a standard therapy, it's not enough
just to
be better than placebo. You really in fact need to
be preserving a substantial fraction, at
least, and
I think what's emerged is, 50 percent of
the
effect.
You need to be preserving that.
So with those as criteria,
where do we go?
What do we do? The analysis that the briefing
document for the sponsor provides, as
well as in
their slide CE 35, basically said let's
estimate
the efficacy, and they called it a
"paper placebo."
Let's estimate the efficacy of Exanta
here by
saying there's two pieces here. We know the
relative risk, even though it's
unfavorable, is
1.39 when you look at Exanta against
warfarin, but
warfarin is so good its relative efficacy
is .36.
If we take the product of those two
pieces we
should be getting a paper estimate of
what the
efficacy is of Exanta against placebo,
and that's
.5, and that comes out to be a very
positive
impression for a result.
There are important strengths
to that
analysis.
First of all, it's based on relative
395
risks instead of absolute differences,
and in this
setting I think that's a more robust way
to analyze
the data.
It also is taking into account the
variability in the estimate of warfarin's
effect.
But there are several issues that
are not
taken into account in that analysis. The first of
them is it's not taking into account the
level of
variability that's occurring in these
estimated
effects across the six warfarin
trials. The FDA,
in attempting to address that, used a
random
effects model to address that issue. It doesn't
take into account the uncertainties
generated by
the fact that four of the six trials
weren't
blinded when you're estimating warfarin's
effect.
Importantly, it doesn't take into account
two other
things.
It doesn't make any accommodation for the
uncertainty about the validity of the
constancy
assumption, and it also doesn't address
the fact
that we have to preserve half of
warfarin's
effects.
So while those analyses looked impressive
at the beginning, there are several
critical issues
that aren't addressed.
396
In essence, what is a proper approach?
Steve talked about dose creep and
duration creep,
so I'll talk about bio-creep.
[Laughter.]
DR. FLEMING: A common term that is used
in the non-inferiority setting is
bio-creep.
What's bio-creep all about? Well, suppose I have a
standard intervention whose efficacy is
reasonably
well understood. Now I go to generation two and I
show I'm not meaningfully worse than
generation one
with non-inferiority. But let's suppose I'm
liberal, like I use a big margin, so my
estimate
actually is I am somewhat worse by
estimate. Well,
Sponsor 3 comes along with a new
product. Which
one are they going to use for non-inferiority? I
think I'll probably choose the second
generation
one, and let's do non-inferiority again
with a
fairly lenient margin. Two or three generations of
this, and what do we really know about
efficacy any
more?
That's bio-creep.
In essence, to address
bio-creep, one has
to have rigorous margins, as the ICH
guidelines
397
clearly indicate. In my words, it's treacherous
when you're implementing a strategy that
can
declare non-inferiority when your actual
point
estimate of the relative efficacy of the
experimental with the active control
actually
favors the active control. If you have such a
lenient margin that you can be estimated
to be
worse, and yet you're still satisfying
non-inferiority, you're setting yourself
up for a
setting of bio-creep.
In essence, the FDA I thought
did a
marvelous job. In fact, my kudos to John Lawrence
and colleagues for really nailing down
all of these
issues.
On page 9 they define several
possible
margins that could be used to try to
address these
issues, and the first point I would argue
is the
margins based on the risk ratio or
relative risk I
think are in fact the most appropriate to
be used
in this setting. These particular approaches that
are defined here not only use the relative
risk at
the bottom half of this table, but
appropriately
398
adjust for the variability in the
estimate of the
active comparitor's effect. They also adjust for
needing to preserve half of the
efficacy. And they
also, by using the random effects model,
adjust it
for the variability.
But they gave you several
choices, and
those several choices are based on two
additional
factors.
Essentially the more liberal choices say
the margin could be in the neighborhood
of 1.56 to
1.65, meaning that you would satisfy
non-inferiority if in a relative risk
sense your
point estimate was sufficiently favorable
that you
could rule out you had a 65 percent
increase or
excess in the rate of stroke. By the way, from a
clinical relevance perspective, how could
you have
a margin bigger than that? I would struggle with
justifying from a patient's perspective
that it's
okay to have stroke rates be more than 65
percent
higher.
I even struggle with whether that in fact
is sufficiently rigorous.
The key point here is that
those two
estimates, 1.56 and 1.65, they're called the
399
Holmgren approach. A variation of this is the
Rothman approach. We see these used at other
advisory committees. These methods do not make any
accommodation for the validity of the
constancy
assumption. So basically you better be darn
confident that the estimate of the
efficacy of
warfarin in those six trials is precisely
accurate
for what the efficacy of warfarin is in
the SPORTIF
trials even though there could be
different
populations, different assessment, and
different
supportive care.
Generally I have great concerns
about
that, especially when you show me the
event rates.
The event rates in warfarin are much
higher in
those six trials than in the SPORTIF V
trial. So I
would far prefer the approaches that are
used that
are called the 95/95 that adjust for the
uncertainty about the constancy
assumption, and
those are margins in the neighborhood of
1.23 to
1.38.
My last point is they differ
only by
whether you really would include all of
the trials,
400
and this is an issue that's always a
struggle.
Which of those six trials are relevant to
the
context in which the SPORTIF V trial was
done?
Maybe none of them fully, but some of them
more so
than others. The one that in particular looks to
be problematic is the EAFT trial that was
done in
patients where you had recent TIAs and
strokes and
you have very high event rates. So if you drop
those out your margin would be 1.23
rather than
1.38.
But if you actually take the
approach that
I only believe the placebo-control
trials, then
your margin is 1. What does that mean, your margin
is 1?
It means you actually have to show
superiority. Unless you show superiority, you
haven't adequately established efficacy.
In closing, where am I on
this? My sense
is I am not sure about whether we can
include all
six or only have to look at the two that
are the
placebo-control trials. My sense of the margin
should be based on the relative risk
estimate here,
and the proper margin is somewhere
between 1 on the
401
conservative side, meaning you have to
show
superiority, up to about 1.4. Something in that
range seems to be aggressive but
potentially
justifiable factoring all of these
issues.
DR. BORER: Let me ask you one additional
question, Tom. I'm inferring from what you said
specifically that the Rothman approach
that Lloyd
showed us, that suggests that you are
likely to
have preserved at least 71 percent of the
warfarin
effect really is not something that you
would
accept as a valid analysis here because
of all the
things you said?
DR. FLEMING: Well, it's a variation of
analyses that are shown in the briefing
document.
It has one plus that I liked, i.e., what
he was
showing today relative to what was in the
briefing
documents. I think was a step in the right
direction of actually addressing needing
to
preserve half of the effect. So that analysis was
getting into that, whereas the analyses
that the
sponsor provided in the briefing document
didn't.
But what that analysis still
didn't do is
402
it still didn't address the variability
that
existed among these six comparitor trials
for
efficacy.
It didn't address the issue of do we
trust all six of those given that only
two of them
are blinded trials. It didn't address, really
importantly from my perspective, there's
no
adjustment made for the constancy
assumption. So
if you're going to trust that estimate
you've got
to be thinking those estimated effects of
warfarin
in those six trials are exactly what
warfarin is
going to do in the SPORTIF V trial in
spite of the
fact that the estimate of the
efficacy--of the
event rate in the warfarin arm in SPORTIF
V was
only about one-third. And actually, if we go back
to Jay Harrow's answer to Steve earlier
today about
why do III and V differ, his answer was,
well,
there's a whole lot of differences in
these
historical control trials of warfarin
effect. I'm
saying don't play that card too hard
because that's
the very issue that makes the validity of
the
constancy assumption.
If you believe that the
difference in the
403
risk level of an event rate on warfarin
is going to
influence warfarin's efficacy--and I
think that's a
valid thing to think--then I can't trust
these six
estimates that come from higher warfarin
background
rates when I'm projecting the efficacy of
warfarin
in SPORTIF V. I don't know if I can, but that's an
issue of concern.
The last thing that he was
doing
today--it's statistically valid but I
really worry
about it--is he was putting together in a
clever
way the pooling of the data from the two
trials,
and the analyses that I'm giving are
basically--what I've been discussing is
looking at
the definition of the margin as you
assess each of
these two trials individually.
So that's some of the
differences or
issues that weren't addressed in that
analysis that
concern me, as I would look at--I would
favor--I
think John Lawrence's analysis was
outstanding. I
think the analysis--basically, I don't
think I've
added a nickel to what he already
did. I just
explained it. I think those analyses lay out the
404
issues in an excellent way and in my
belief you
have to address the concerns that he
raised in his
summary.
DR. BORER: Can I ask, do you want us
actually to vote on this issue or can we
just go on
to the next question?
[Laughter.]
DR. PICKERING: Could I ask a question?
DR. BORER: Yes, Tom.
DR. PICKERING: My understanding is that
the analyses basically ignored SPORTIF
III and said
let's just focus on SPORTIF V. So my question is,
if you included SPORTIF III how much does
this
change these errors of margin, because
obviously it
went in a different direction?
DR. FLEMING: I was sticking to Question
9, which basically just said, formulate
the margin
of what would be an appropriate margin in
this
setting.
Question 10 gets into part of what you're
talking about in my view, which is, all
right, with
this as a background, how do we assess
the strength
of evidence in SPORTIF III and SPORTIF
V? So I was
405
going to get into that when I got to
Question 10
DR. BORER: Steve?
DR. NISSEN: I just had to make one
comment.
First of all, that was fabulous, and I
always learn a lot from you. But there also is a
clinical context, and all of that
assumes, it's all
oriented around the issue of efficacy,
and in
deciding on a design for the future,
which is
really partly what we're talking about,
is you
can't separate efficacy and safety
because the
efficacy question has a safety component,
namely
that if two therapies have equal safety,
you did a
great job of telling us how to set the
margins for
non-inferiority for efficacy. But that to me isn't
the question that we now face, that the
standard
would have to be modified for what would
constitute
adequate evidence of efficacy if we know
certain
things about adverse safety
consequences. I don't
know if I'm making sense to you or not,
but I think
you can't--I think it was an absolutely
right on if
you had two therapies that were equally
safe.
DR. FLEMING: I believe I follow what
406
you're saying, Steve, and if so, I agree
with you,
and it was one of the first things that I
had
mentioned, as I said what are the
criteria I would
look to in defining a margin? The first criterion
that I would look to is: from a clinical
perspective, what level of excess risk
could be
allowable before to a patient this would
be an
unacceptable level of excess? That should be an
important criterion. And that does need a
subjective assessment of what are the
other
elements of the benefit to risk profile
that this
intervention provides?
If I was looking at
thrombolytic and I
could get rid of intracranial hemorrhage
risk, I
might have a little more of a margin I'd
allow on
mortality in that type of setting. If there is in
this setting a judgment that the level of
safety
profile is substantially enhanced, I can
understand
a willingness to allow a somewhat larger
element to
the margin.
So there are two separate
domains that
really should define the margin. One is a clinical
407
relevance aspect that is really a
rigorous fair
assessment of how much worse can this be
in the
efficacy measure in the context of
overall benefit
to risk?
And the other is very much a statistical
issue of what is the scientific evidence
that we
have to truly allow us to conclude that
we are
effective when we don't have the luxury
of doing a
direct comparison in the randomized
trial? We're
having to rely on the historical
data. There are
two pieces, and I always say a chain is
as strong
as its weakest link. The weakest link in most
non-inferiority analyses is the uncertainty
about
the strength and the reliability of the
evidence of
the active comparitor against the
placebo.
DR. BORER: We'll move on to the final
question, which is: Based on the currently
available data do you conclude that the
benefits of
ximelagatran for long-term use for
prevention of
stroke and systemic embolic events in
patients with
atrial fibrillation outweigh its risk?
Steve, you can go ahead and
start.
DR. NISSEN: Again, I'm going to be
408
short-winded rather than long-winded
because I
think people want to catch planes.
You know, I think I've made it
pretty
clear that I think it's no, and--
DR. HOUN: Can I just interrupt and let
the two departing members vote on the
record?
DR. BORER: They did.
They both voted no.
DR. CUNNINGHAM: Susanna Cunningham, no.
DR. HIATT: No.
William Hiatt votes no.
DR. NISSEN: And to directly answer Tom's
question about where to set that margin
if this
development program continues, the margin
is 1.0.
In
other words, I'm setting that margin, knowing
what I know about safety, that at a very
minimum
that a future trial would have to show
superiority
in order to compensate for what I think
is a pretty
well-defined increase in risk of a very
serious
toxicity.
DR. BORER: Tom?
DR. PICKERING: I would say yes. Viewing
it in the sort of relatively narrow
confines of the
trials, but looking at the greater
population of
409
patients who currently are not getting
any
treatment and who have a prospect of
getting
treated if this drug is approved, I would
say the
benefits--there are substantial benefits
that
outweigh the risks.
DR. PORTMAN: No, for the same reasons as
the last long-term trial.
DR. BORER: Beverly?
DR. LORELL: No.
DR. SACKNER-BERNSTEIN: I would say no
also, realizing that there seems as
though there
are a number of patients whose long-term
status in
the SPORTIF trials are unknown because
they
withdrew from the study, not just
withdrew from
study medication. So, no.
DR. BORER: Paul, do you have any comment
to make?
No.
DR. BORER: Okay.
I reluctantly vote no,
primarily because of the arguments that
Tom has
made.
Like Tom, I believe that there's a place for
this drug in this population. I think we need to
know more about the risk. We need to define it
410
better, and we need to be able to
minimize it
better than we have, just as I said about
the other
long-term issue. So I agree with Steve about the
risk/benefit concern here. But I'm primarily
concerned that I'm not secure that we
know that
we've actually preserved the degree of
warfarin-related benefit that we nominally
think we
did from this study. I'd like some reassurance on
that, and we may not get that without
another
trial.
Having said that, I don't think
that we
need, as I said earlier in another
context, that we
need to show superiority. I think there are enough
good reasons why an alternative to
warfarin would
be a good thing so that if we understood
the risk
and we could minimize it by some reasonable
algorithm in a way that would be
acceptable to
us--and I can't define that here--that
equivalence
would be sufficient if we could be sure
it was
equivalent. So I would vote no on that basis.
DR. TEERLINK: I share your reluctance to
vote no.
I think it's a potentially very useful
411
therapy in this patient population and
would help a
great deal of patients who are not being
helped at
this time. But my concerns about the safety remain
unaddressed at this time, as well as the
need to
show true equivalence to warfarin. I think I'm
probably between you two, Steve and Jeff,
that I
don't know if I'd make it 1.0 and I don't
know
where I would put that line. It would have to be
fairly stringent.
DR. BORER: Dr. Sjogren?
DR. SJOGREN: I agree with you and John,
and I vote no.
DR. BORER: Alan?
DR. HIRSCH: The reluctant no again, but
with a comment. There is a pressing need for such
a therapeutic molecule as you've said,
and the goal
to decrease both stroke as well as DVT
and
pulmonary embolus is required with such
molecules,
but this database for this particular
medication
doesn't permit me to vote yes quite yet.
The addition of a new molecule
doesn't
mean that treatment intensity would have
increased
412
for the public health advocates. That's a function
of education, guidelines and care
management
pathways, not drug approval, per se.
DR. BORER:
I think we've answered all the
questions. Oh, Tom, didn't you vote?
DR. FLEMING: I don't think so, but I'll
be brief.
DR. BORER: Sorry.
DR. FLEMING: I will be brief on this. I
look at these two pivotal--well, I look
at these
two studies. I look at V as being the pivotal
study and III as being the supportive
trial.
Largely with the blinding and open label
rationale
for that, but also the V study done in
the U.S. I
think is particularly relevant to a U.S.
regulatory
action.
For the reasons that I
discussed, I don't
think, in fact, I believe quite clearly
that the
SPORTIF V trial did not establish efficacy. The
SPORTIF III trial, even though it is open
label and
has that (?)-arity, so to speak, it has a
favorable
estimate.
I'm uncertain that I would require
413
superiority, but I do believe that I would require
nothing more lenient than an upper limit
on the
order of 1.4 or so. That would, however, be a
criterion that the SPORTIF III trial does
meet, and
so ultimately whether SPORTIF III is
viewed as
positive is influenced by how important
is
blinding, the blinding issue in
assessment of
reliability of results? How important is it that
four of the six warfarin trials weren't
blinded?
And how important is it that the
uncertainties
about the hepatic toxicity makes, as some
have
argued, one to want to be much more
stringent in
terms of how you would look at this.
Ultimately though, it's a
strength of
evidence issue as well. I want SOE 2, the strength
of evidence of two trials, and at best
here I'd say
we have one. So that's the rationale for my view
of no.
DR. BORER: I was off by 26.5 minutes, but
I think we've answered all the
questions. Thank
you very much.
DR. NISSEN: Jeff, can I just say one more
414
thing unrelated?
You know, this is your last
meeting as
chair, and I guess I wanted to express,
on behalf
of a lot of us who have served with you
over many
years, appreciation for the great
leadership and
balance you have provided to the
Committee, and I
think everybody--that all the
stakeholders here owe
you a debt of gratitude for the fair and
evenhanded
way you've run this Committee over the
years I've
certainly been involved. So my thanks.
[Applause.]
DR. BORER: Thank you very much. And
we'll close the meeting.
[Whereupon, at 5:28 p.m., the
Advisory
Committee was adjourned.]
- - -