1
DEPARTMENT OF HEALTH AND HUMAN
SERVICES
FOOD AND DRUG
ADMINISTRATION
CENTER FOR DRUG EVALUATION AND
RESEARCH
ONCOLOGIC DRUG ADVISORY
COMMITTEE
Hilton
2
PARTICIPANTS
Donna Przepiorka, M.D., Ph.D., Chair
Johanna M. Clifford, M.S., RN,
Executive Secretary
MEMBERS:
John T. Carpenter, Jr., M.D.
Bruce D. Cheson, M.D.
James H. Doroshow, M.D.
Stephen L. George, Ph.D.
Antonio J. Grillo-Lopez,
M.D.
Pamela
J. Haylock, RN
Silvana Martino, D.O.
Gregory
H. Reaman, M.D.
Bruce G. Redman, D.O.
Maria Rodriguez, M.D.
Sarah A. Taylor, M.D.
CONSULTANTS (VOTING):
Michael Bishop, M.D.
Ronald Bukowski, M.D.
Ralph D'Agostino, Ph.D.
Maha Hussain, M.D.
Jan Buckner, M.D.
Wen-Jen Hwu, M.D.
Joanne Mortimer, M.D.
Michael Perry, M.D.
PATIENT REPRESENTATIVES (VOTING):
Kenneth McDonough (for Genasense)
Natalie Compagni-Portis
(for RSR 13 Injection)
FDA STAFF:
Richard Pazdur, M.D.
Grant Williams, M.D.
Robert Temple, M.D.
3
C O N T E N T S
Opening Remarks, Donna Przepiorka, M.D., Ph.D. 5
Comments by Congressman Peter
Deutsch 5
Comments by Alex Delpizo 11
Conflict of Interest Statement,
Johanna M. Clifford, M.S., RN 14
Opening Remarks, Richard Pazdur,
M.D. 20
Genta Presentation:
Introduction, Loretta M. Itri,
M.D. 26
Melanoma Overview, John Kirkwood,
M.D. 29
Study GM301, Loretta M. Itri,
M.D. 36
Clinical Benefit Summary,
Frank Haluksa, M.D., Ph.D. 60
FDA Presentation:
Medical Review, Robert Kane, M.D. 69
Statistical Review, Peiling Yang,
Ph.D. 76
Clinical Relevance, Robert Kane,
M.D. 86
Questions from the Committee 93
Open Public Hearing 125
Committee Discussion 152
Allos Presentation:
Introduction, Pablo J. Cagnoni,
M.D. 206
Brain Metastases, John H. Suh,
M.D. 209
The Science of RSR13, Biran D.
Kavanaugh,
M.D., MPH 216
Clinical Efficacy Results,
Pablo
J. Cagnoni, M.D.
225
Conclusions, Paul A.
Bunn, Jr., M.D. 251
FDA Presentation:
Clinical Review, Kevin Ridenhour,
M.D. 254
Statistical Review, Rajeshwari
Sridhara, Ph.D. 265
4
C O N T E N T S
Questions to the FDA and the Sponsor 278
Open Public Hearing 309
Subgroup Analysis in Clinical Trials,
Stephen George, Ph.D. 314
Committee Discussion 333
5
1 P R O C E E D I N G S
2 Opening Remarks
3 DR. PRZEPIORKA: Good morning to all and
4
welcome to the Food and Drug Administration's
5
Advisory Committee for Oncologic Drugs.
My name is
6
Donna Przepiorka. I will be
chairing the
7
committee. I just wanted to
remind everyone in the
8
audience that the purpose of the individuals on
9
this panel is to serve as independent consultants
10 to
the FDA. We do not work for the
FDA. We are
11
also not anyone who makes any decisions; we only
12
provide advice.
13
Our first item on the agenda--we are going
14 to
go a little bit out of order. We want to
hear
15
first from Congressman Deutsch who has a few words
16 to
say.
17
CONGRESSMAN DEUTSCH: Thank you
very much.
18 I
appreciate the opportunity to be here.
My name
19 is
Congressman Peter Deutsch, and I recognize that
20 it
is not at every meeting of this committee that
21 you
are addressed by a member of Congress.
Largely
22 it
is in that capacity that I speak to you today,
6
1 but
it is also in my capacity as an individual who
2 has
been personally affected by the specter of
3
melanoma.
4
On several occasions I have had basal
5
cells removed from my body.
Thankfully, they were
6 not
malignant but their existence renders me high
7
risk. My dermatologist now
evaluates me on a
8
quarterly basis for melanoma and guides me on how
9 to
reduce my risk profile. I pray that this
risk
10
never materializes but, if it does, I need to know
11
that my physician and I have access to every
12
therapeutic treatment available for this horrible
13
disease. As someone who actually
hears people
14
testify in many settings, I am trying to get your
15
attention so actually I have pictures of my kids
16 who
both have red hair so, obviously, they are high
17
risk for skin cancer as well especially as having a
18
parent who has been diagnosed with basal cells.
19
They also happen to live in Florida.
20
Again, most of the people in this room
21
don't live in Florida and I am not exaggerating
22
that the school that they go to and, in fact, the
7
1
schools they have gone to since pre-K, do not have
2
hallways. It is one of the unique
things about
3
Florida, south Florida in particular so they are
4
literally outside all the time.
For anyone who has
5
kids, especially in a setting like south Florida,
6
think about the summer when you try to get your
7
kids to wear suntan lotion. It is
not an easy
8
thing to do. So, this is a very
real thing. I
9
mean, I have fights with my kids, especially as
10
they have gotten older, about putting suntan lotion
11 on,
on a continuous basis.
12
But it is not just for my kids; it is not
13 for
myself that I am here today. It is for
all the
14
constituents I represent and all the citizens
15
around the nation. So, it is on
their behalf as
16
well that I stand before you today, not to advocate
17 for
the approval of this drug but to advocate that
18 the
mind set from which you consider this
19
application be your own mind set--clinical
20
physicians dedicated to the welfare of their
21
patients.
22
What does this mean? That this
8
1
application be a referendum on whether you would
2
want this drug available to your patients if they
3
were diagnosed with metastatic melanoma.
That is
4 the
standard we owe cancer patients and that is the
5
standard government is obligated to uphold.
6
I did not come here to preach to this
7
committee to the extent me and Congress have had
8
frustration with over-regulation by the FDA. It is
9 not
of your doing; quite the opposite. It is
10
people like yourselves who give up your time to
11
guide the FDA. I cannot
over-emphasize the
12
importance of your role. You
provide the FDA a
13
window that they otherwise do not have, a window
14
into the real world, if you will, a world in which
15
dying cancer patients are desperate for and must be
16
given access to every reasonable treatment that
17
might save their lives.
18
As you may know, there were two relevant
19
newspaper articles last week that got some
20
attention in Congress. One was an
article in The
21 New
York Times about a Japanese study published in
22 The
New England Journal of Medicine proving the
9
1
effectiveness of a drug called UFT in treating a
2
form of lung cancer. What was
staggering about the
3
article was that this same technology was rejected
4 in
this country by the FDA. In other words,
5
thousands of cancer patients in this country could
6 be
dying because the government failed them.
7
What I later learned was that the FDA
8
rejected this drug even though this very advisory
9
committee composed of your predecessors voted
10
unanimously to approve it and, because the FDA did
11 not
accept the recommendations of clinicians,
12
countless Americans lack access to that drug today.
13
That is inexcusable.
14
In the other article, the Wall Street
15
Journal related to this committee's hearings. It
16
offered no views on whether this drug should be
17
approved but, instead, noted the absence of
18
treatments for metastatic melanoma and a couple of
19
vignettes about the people who took the drug. One
20 of
those was an individual names David Bernstein
21 who
is scheduled to join us here today. Mr.
22
Bernstein is a fourth grade teacher from a small
10
1
town in New Jersey. The article
said that Mr.
2
Bernstein's cancer went away and he is alive today,
3
teaching his students in his fourth grade classroom
4
because of the drug before you today.
5 I am not a physician nor a scientist
and I
6
have not studied the clinical data regarding this
7
drug, but I do know this, if you find that this
8
drug is as safe and effective as other available
9
treatments, if it reasonably presents another
10
possible course of treatment, by what right can
11
government deny cancer patients an avenue to save
12
their lives? This is not about a
passing illness
13 for
which there are other treatments. This
is
14
about cancer, an absolutely devastating disease
15
that has in some ways affected nearly every single
16
American. This is about cancer
patients who are
17
dying and desperate for a chance to live longer.
18 It
is in their interest that we must be foremost in
19
today's hearing.
20
I flew back to Washington last night to
21
speak to you this morning, however, prior
22
obligations in my district require me to actually
11
1
literally turn around right now and return to
2
Florida this morning. I regret
that I can't stay
3
here to listen to all of the testimony but I wish
4 to
thank this committee for its time, and it has
5
been an honor and pleasure to speak with you this
6
morning.
7
DR. PRZEPIORKA: Thank you,
Congressman
8
Deutsch. Any questions for the
Congressman?
9
[No response]
10
Thank you, sir.
11
CONGRESSMAN DEUTSCH: Thank you.
12
DR. PRZEPIORKA: Next we will hear
from a
13
representative from Congressman Ferguson's office.
14
MR. DELPIZO: My name is Alex
Delpizo. I
15 am
here representing Congressman Mike Ferguson of
16 New
Jersey who, unfortunately, is in New Jersey and
17
couldn't be here with us today.
18
I am not a scientist or a clinician or a
19
chemist but everyone knows a person whose life has
20
been taken by cancer. For me,
that person was my
21
mother. She fought and eventually
lost her
22
six-year battle with cancer.
However, due to
12
1
miracle life-extending drugs she saw two of her
2
children get married and met her three
3
grandchildren. My mother was
fortunate enough to
4
experience all of the wonderful things that mothers
5 and
grandmothers experience later in life.
6
As you know, Genasense us used to treat
7
stage 4 metastatic melanoma.
Metastatic melanoma
8 is
currently a death sentence. When two
available
9
therapies treat the disease and the last
10
chemotherapy therapy treatment was approved in
11
1975, yours is an awesome responsibility. The FDA
12
works every day to ensure that Americans and their
13
food and drug supply are safe.
Your decisions on
14
which drugs are approved are based on numbers, and
15
numbers are very important, however, we would never
16 want to approve a placebo. However, an
17
over-emphasis on statistics at the expense of
18
patient needs does a life-threatening disservice.
19 The
failure to appreciate mean or median
20
statistical analyses in any size sampling also
21
fails to take into account a patient population
22
that achieved the most dramatic overall response.
13
1
Given the devastating nature of this
2
disease and the relatively few treatments
3
available, even marginal increases in life
4
expectancy can clearly be the difference between
5
rapid death and years of life extension for those
6
patients that will see a benefit from this and
7
other drugs.
8
In closing, I would like to highlight the
9
experience of one of my constituents in Montgomery
10
Township in New Jersey. David
Bernstein was
11
diagnosed with skin cancer and prescribed
12
chemotherapy to remove a grape-sized tumor on his
13
chest. Mr. Bernstein opted to
supplement the
14
chemotherapy by joining a clinical trial of an
15
experimental drug. Six weeks
after his first dose
16 he
received the news that his tumor had essentially
17 disappeared.
This was two years ago. That
18
experimental drug was Genasense.
19
For my mother, David Bernstein and for all
20 of
those who have been diagnosed with cancer, I
21
respectfully request that you look favorably on
22
Genasense and other new drug applications that can
14
1
provide hope for those for whom hope is all they
2
have. Thank you very much.
3
DR. PRZEPIORKA: Thank you. Again, I
4
would like to ask the folks who are standing along
5
that far wall by the doors to please step outside
6
into the hall, or take a seat, or take a stand at
7 the
back wall only, please. You are going to
need
8 to
vacate that area immediately, please.
9
We would like to now move on to the first
10
item on the agenda and Johanna Clifford will read
11 the
conflict of interest statement. Thank
you.
12 Conflict of Interest Statement
13
MS. CLIFFORD: Thank you. The following
14
announcement addresses the issue of conflict of
15
interest with respect to this meeting and is made a
16
part of the record to preclude even the appearance
17 of
such at this meeting.
18
Based on the submitted agenda and
19
information provided by the participants, the
20
agency has determined that all reported interests
21 in
firms regulated by the Center for Drug
22
Evaluation and Research present no potential for a
15
1
conflict of interest at this meeting, with the
2
following exceptions:
3
In accordance with 18 USC Section
4
208(b)(3), Dr. Ronald Bukowski has been granted a
5
waiver for serving on a competitor's advisory board
6 on
an unrelated matter for which he receives less
7
than $10,000 a year; consulting with the sponsor of
8
dacarbazine on an unrelated matter for which he
9
receives less than $10,000 a year; and, finally,
10 for
consulting with a competitor on an unrelated
11
matter for which he receives less than $10,000 a
12
year.
13
Dr. Maha Hussain has been granted waivers
14
under 18 USC 208(b)(3) and 21 USC 505(n) for
15
unrelated consulting for the co-developed of
16
Genasense for which she receives less than $10,000
17 a
year; and owning stock in the co-developer of
18
Genasense, valued from $25,001 to $50,000.
19
Dr. Wen-Jen Hwu has been granted a limited
20
waiver under 18 USC 208(b)(3) for her employer's
21
contract with a competitor for an
22
investigator-initiated study of a competing
16
1
product. The contrast is less
than $100,000 a
2
year. Under the terms of the
waiver, Dr. Hwu will
3 be
permitted to participate in the committee's
4
discussions of Genasense. She
will not, however,
5 be
able to vote.
6
A copy of these waiver statements may be
7
obtained by submitting a written request to the
8
agency's Freedom of Information Office, Room 12A-30
9 of
the Parklawn Building.
10
We would also like to disclose that Dr.
11
Silvana Martino has been recused from participating
12 in
all matters concerning Genta's Genasense.
13
Lastly, we would like to note for the
14
record that Dr. Antonio Grillo-Lopez, Chairman,
15
Neoplastic and Autoimmune Diseases Research
16
Institute, is participating in this meeting as in
17
industry representative, acting on behalf of
18
regulated industry. He would like
to disclose that
19 he
is a scientific advisor to Chiron and receives
20
speakers fees from Roche.
21
In the event that the discussions involve
22 any
other products or firms not already on the
17
1
agenda for which FDA participants have a financial
2
interest, the participants are aware of the need to
3
exclude themselves from such involvement and their
4
exclusion will be noted for the record.
5
With respect to all other participants, we
6 ask
in the interest of fairness that they address
7 any
current or previous financial involvement with
8 any
firm whose product they may wish to comment
9
upon.
10
DR. PRZEPIORKA: Thank you. Once again,
11
there are still some folks registered for the open
12
public hearing who have not signed in.
I just want
13 to
remind you that if you do wish to speak at the
14
open public hearing you will need to sign in at the
15
table outside.
16
Next, I would like the members of the
17
committee and the other participants to introduce
18
themselves and we will start with Dr. Pazdur.
19
DR. PAZDUR: Richard Pazdur,
Director of
20 the
Division of Oncology Drug Products, FDA.
21 DR. WILLIAMS: Grant Williams, FDA,
22
Director, Division of Oncology Drugs.
18
1
DR. FARRELL: Ann Farrell,
clinical team
2
leader for Genasense.
3 DR. KANE: Robert Kane, medical reviewer.
4
DR. YANG: Peiling Yang,
statistical
5
reviewer.
6
DR. BUKOWSKI: Ron Bukowski,
medical
7
oncologist, Cleveland.
8
DR. BISHOP: Michael Bishop, Experimental
9
Transplantation, Immunology Branch, National Cancer
10
Institute.
11
DR. HWU: Wen-Jen Hwu, medical
oncologist
12 at
the Memorial Sloan-Kettering.
13
DR. TAYLOR: Sarah Taylor,
University of
14
Kansas.
15
DR. REAMAN: Gregory Reaman,
George
16
Washington University and Children's National
17
Medical Center.
18
DR. REDMAN: Bruce Redman,
University of
19
Michigan.
20
MS. CLIFFORD: Johanna Clifford, FDA,
21
executive secretary for this meeting.
22
DR. PRZEPIORKA: Donna Przepiorka,
19
1
University of Tennessee, Memphis.
2
DR. RODRIGUEZ: Maria Rodriguez,
medical
3
oncologist, M.D. Anderson Cancer Center.
4
DR. DOROSHOW: Jim Doroshow,
Division of
5
Cancer Treatment and Diagnosis, NCI.
6
DR. CHESON: Bruce Cheson,
Georgetown
7
University Lombardi Comprehensive Cancer Center.
8
DR. GEORGE: Stephen George, Duke
9
University.
10
MS. HAYLOCK: Pamela Haylock. I am a
11
nurse and I am at the University of Texas.
12
DR. CARPENTER: John Carpenter,
University
13 of
Alabama at Birmingham.
14
DR. D'AGOSTINO: Ralph D'Agostino,
Boston
15
University biostatistician.
16
DR. MORTIMER: Joanne Mortimer,
medical
17
oncology Eastern Virginia Medical School.
18
DR. HUSSAIN: Maha Hussain,
University of
19
Michigan.
20
MR. MCDONOUGH: Ken McDonough,
patient
21
representative.
22
DR. GRILLO-LOPEZ: Antonio
Grillo-Lopez,
20
1
Neoplastic and Autoimmune Diseases Research
2
Institute.
3
DR. PRZEPIORKA: Thank you to
all. I
4
think Dr. Pazdur will open with some remarks.
5 Opening Remarks
6
DR. PAZDUR: Thank you very much,
Donna.
7
First, I would like to recognize the contributions
8 of
four ODAC members who will be leaving the
9
committee after this meeting.
These members
10
include our chairman, Donna Przepiorka, John
11
Carpenter, Sarah Taylor and Bruce Redman. We, at
12 the
FDA, recognize their efforts at providing us
13
advice at these public meetings and, in addition,
14 we
appreciate their valuable assistance throughout
15 the
years in providing us with their insights at
16
other FDA meetings and in reviewing and assessing
17
protocols. Our work and the
welfare of the
18
American public is greatly facilitated by their
19
hours of work and their talents devoted to these
20
tasks. Again, Donna, John, Sarah
and Bruce, we
21
thank you for your efforts, your patience with our
22
phone calls, and advice on some of the most
21
1
perplexing issues of drug development.
Let me say
2
this, this is not "adios" but "hasta la vista" and
3 it
is not "hasta la vista, baby."
We will be
4
calling you; we will be in touch; this will be a
5
continuous process that we will be dealing with you
6
over the years, but we do appreciate your kindness
7 and
your efforts at helping us with some of the
8
problems that we have at hand.
9
Let's turn to the issues at hand.
This
10
morning's meeting focuses on a drug for the
11
treatment of patients with advanced melanoma who
12
have not received prior chemotherapy.
I would like
13 to
spend some time addressing issues for you to
14
consider during the presentations provided by the
15
sponsor and the FDA staff. These
issues are
16
important to this application but also this
17
afternoon's application and in drug development in
18
general, especially as we have continuing, ongoing
19
discussions and dialogue with the committee on
20
endpoints for drug development.
21
The FDA has long considered the
22
demonstration of an improved survival as the gold
22
1
standard for drug approval. An
improvement in
2
survival associated with an acceptable safety
3
profile is of unquestionable clinical benefit. It
4 is
assessed daily and is unambiguous. When
we, at
5 the
FDA, began our discussions with the committee
6 on
drug approval we realized that there may be some
7
disadvantages to requiring survival improvement for
8
drug approval. These
disadvantages include the
9
confounding of survival analysis by crossover with
10
frequently large patient numbers required to be
11
enrolled on trials for survival, and the long
12
follow-up that may be required in selected
13
oncological diseases.
14
This trial at hand this morning was
15
originally discussed with the agency to be a trial
16
with a primary endpoint of survival improvement.
17 The
trial did not demonstrate an improvement in
18
overall survival. We are asked to
evaluate this
19
drug for approval on the basis of secondary
20
endpoints of claimed improvements in
21
progression-free survival or PFS and response
22
rates. Please member that since
this drug is added
23
1 to
a standard therapy we must assess the drug's
2
contribution to that standard therapy and any
3
claimed response rates or claims for PFS advantages
4
represent a combination of the investigational
5
agent and the standard therapy.
Hence, we must
6
isolate the efficacy of the drug in assessing the
7
drug's efficacy.
8
Let's turn our attention to the
9
measurement and assessment of PFS which will be
10
discussed during this meeting on multiple
11
occasions. The assessment of PFS
may be difficult
12 and
uncertain in unblinded trials with a small
13
effect on this endpoint and where there is a lack
14 of
attention to clinical trial issues that are
15
important in measuring and comparing PFS data
16
between treatment arms. These
issues include a
17
prospectively defined methodology for assessing,
18
measuring and analyzing PFS.
These need to be
19
detailed in the protocol and in the statistical
20
plan. Tumor progression should be
carefully
21
defined in the protocol. The FDA
and the sponsor
22 should agree prospectively on the protocol,
the
24
1
case report forms and the statistical analysis plan
2 for
PFS. There should be a prespecified
analysis
3
plan for handling missing data, especially missed
4
assessment visits. Censoring
methods and
5
assessment of progression in non-measurable lesions
6
must be prospectively outlined and agreed upon.
7
Most importantly, visits and radiological
8 assessments
should be symmetrical on the study arms
9 to
prevent systematic bias. When possible,
studies
10
should be blinded. This is
especially important
11
when the patient or investigator assessments are
12
included as components of the progression endpoint.
13 If
progression is assessed by both the treating
14
physician and an external review panel or an
15
external radiology committee, the protocol should
16
prospectively stipulate whose assessment will be
17 used
in defining PFS. This cannot occur after
the
18
study data has been examined.
19
Hence, from a practical perspective, PFS
20 as
a primary endpoint for drug approval takes
21
meticulous, prospective planning.
The measurement
22 of
PFS progression-free survival requires rigor.
25
1
This planning is frequently lacking in clinical
2
trials that relegate PFS to a secondary endpoint.
3
Some practical problems outlined above in
4
accurately characterizing the treatment of PFS will
5 be
discussed by the FDA reviewers.
6
Provided an acceptable safety profile, one
7 has
to answer the following question, what is the
8 magnitude
of the drug's effect on PFS that would be
9
considered clinically relevant? A
very small
10
effect may raise questions about the very existence
11 of
this effect, especially when the study is
12
unblinded and attention to the symmetry of
13
assessments and handling of missing assessments is
14 not
evident.
15
In answering whether marketing approval
16
should be granted to an agent, two important
17
questions need to be answered.
First, does the
18 drug have a convincing effect that can be
19
adequately characterized?
Secondly, and this
20
question can only be addressed if the first
21
question is answered in the affirmative, what is
22 the
clinical relevance of the effect? This
26
1
obviously must take into account a risk-benefit
2
analysis. However, benefit can
only be assessed in
3
this equation if it convincingly exists and also
4 can
be adequately characterized.
5
I hope these comments will provide a
6
catalyst for your considerations this morning, this
7
afternoon and tomorrow as we discuss endpoints of
8
drug approval. Donna, I turn the
program over to
9 you and I will answer questions after the FDA
10
presentations. Thank you.
11
DR. PRZEPIORKA: Thank you, Dr.
Pazdur.
12
Let's go ahead and begin with the sponsor
13
presentation, with an introduction by Dr. Itri.
14
Sponsor Presentation
15 Introduction
16
[Slide]
17
DR. ITRI: Dr. Przepiorka, members
of the
18
Oncology Drug Advisory Committee, ladies and
19
gentlemen, it is my pleasure, on behalf of Genta,
20 to
introduce the agenda and the participants for
21 the
presentation of the new drug application for
22
Genasense in combination with dacarbazine for the
27
1 treatment of patients with advanced
malignant
2
melanoma.
3
Following my introductory remarks, Dr.
4
John Kirkwood will give an overview of malignant
5
melanoma and available treatments.
After Dr.
6
Kirkwood's presentation I will return to the podium
7 and
discuss the results of GM301 in detail.
At
8
that point, Dr. Frank Haluska will summarize the
9
risks and benefits in the context of the disease we
10 are
treating.
11
[Slide]
12
By way of introducing our speakers, Dr.
13
Frank Haluska is from Harvard University and Mass.
14
General Hospital. He is chairman
of the CALGB
15
melanoma committee. Dr. John
Kirkwood is professor
16 and
vice chairman of Medicine at the University of
17
Pittsburgh and is also chairman of the ECOG
18
melanoma committee.
19
[Slide]
20
In addition to our distinguished speakers,
21 we
are fortunate to have with us today a number of
22 clinical
experts in the field of melanoma,
28
1
including Dr. Sanjiv Agarwala from the University
2 of
Pittsburgh Cancer Center, Dr. Agop Bedikian from
3
M.D. Anderson Cancer Center, Dr. Paul Chapman from
4 the
Memorial Sloan-Kettering Cancer Center, Dr.
5
Robert Conry from the University of Alabama, Dr.
6
Peter Hersey from the University of Newcastle, all
7 the
way from Australia, and Dr. Evan Hersh from the
8
University of Arizona Cancer Center.
9
Drs. Bedikian, Conry, Hersey and Hersh
10
were principal investigators in our study and
11
together are responsible for managing approximately
12 20
percent of patients who are on our trial.
They
13 are
available to address any issues you may have
14
regarding patient management in the study. Dr.
15
Janet Wittes, formerly head of statistics at the
16
National Heart, Lung and Blood Institute and
17
currently president of Statistics Collaborative, is
18
available to provide expert biostatistical
19
consultation. Dr. Robert Ford,
chief medical
20
officer and founder of RadPharm, is with us to
21
address the intricacies related to the blinded
22 independent
review of radiographic studies. I
29
1
would like to now invite Dr. John Kirkwood to the
2
podium.
3 Melanoma Overview
4
DR. KIRKWOOD: Thank you, Loretta.
5
[Slide]
6
Dr. Pazdur, Dr. Przepiorka, members of
7
ODAC and the FDA, I am delighted to speak with you
8
today about a disease that many of us here have
9
spent all of our lives working on.
10
[Slide]
11
This is a disease that has risen in
12
epidemic proportions and is 4 percent of new
13
cancers, rising at 5 percent per year.
The
14
mortality from this cancer is also rising and most
15 notably
for men over 50 for whom there is a 157
16
percent increase in mortality in just the last
17
decade. The societal impact of
this cancer is even
18
more because of its median age of incidence in the
19
late 40s, and it takes a toll in terms of
20
productive life years that exceeds many more
21
frequent cancers, even including prostate cancer.
22
[Slide]
30
1
In the past 37 years only three agents
2
have been approved for the treatment of this
3
disease in the advanced setting.
Not one of these
4
agents was approved on the basis of randomized,
5
controlled Phase 3 trials prior to their approval.
6
None of these agents has ever shown a survival
7
benefit. Approval of these agents
was based solely
8 on
response rate.
9
Hydroxyurea, approved in 1967 with a 10
10
percent response rate, has not been used in the
11
clinical community for 20 years or more.
12
Dacarbazine, approved in 1975 with a
13
response rate of 23-25 percent, has more recently
14
been summarized in an article to appear next month
15 in
the European Journal of Cancer. The
response
16 rates that range between 7-13 percent I think
are
17 far
more accurate assessments of the true response
18
rate to this agent. Most of these
were done
19
pre-RECIST criteria and we don't know really what
20 the
objective response rate will be in larger
21
trials using the newer RECIST criteria that have
22
been used for the study to be discussed today.
31
1
[Slide]
2
Turning to IL-2, the most recent agent
3
approved for the treatment of metastatic melanoma,
4 the
IL-2 NDA pooled 8 Phase 2 small studies.
The
5
regimen was not compared in these to any other
6
therapy. The approval was based
upon quality of
7
response, durable responses and, given the
8
significant toxicity of this agent, the population
9
that was treated was highly atypical of the general
10
community of patients that we have to deal with in
11 the
country at large. The median age was 42
years.
12 The
patients had in general no co-morbidity in
13
terms of cardiac or pulmonary disease.
Most of the
14
patients who had responses had disease confined to
15
skin, lymph nodes and lung. The
toxicity of this
16
regimen is so regularly, predictably severe that,
17 in
fact, specialized units are required for the
18
administration of this agent. Its
administration
19 is
confined to specialized centers in general
20
across the country.
21
[Slide]
22
IL-2 responses were noted in 16 percent of
32
1
patients treated, about one-third of whom had
2
surgery to maintain this complete response, and 10
3 percent
partial responses, defined using pre-RECIST
4
criteria. The most salient aspect
of the IL-2
5
benefit in these patients has been the long
6
duration of response observed in some patients.
7
While the median duration of patients treated at
8
large was 9 months, the median duration for
9
patients who achieved complete responses was
10
greater than 5 years.
Unfortunately, the number of
11
those complete responses alive is rather small.
12 The
drug-related mortality with this treatment in
13
this series was 2 percent, further compromising
14
this relative benefit.
15
[Slide]
16
Over the years there have been many
17
attempts to improve upon the therapeutic benefit of
18 dacarbazine.
The largest of the trials conducted
19 in
the last five years are summarized in this
20
slide, beginning with the IL-2 experience which was
21
Phase 2 and, therefore, for which no comparator
22
exists.
33
1
These include the Dartmouth regimen,
2
adding tamoxafin to BCNU, cisplatin and
3
dacarbazine; two regimens of biochemotherapy
4
including one that the Eastern Cooperative Oncology
5
Group and the Intergroup presented to the ASCO
6
meetings just a year ago, now enrolling 416
7
patients; and a similarly large study from the
8
EORTC that has not yet been published; as well as a
9
publication just recently in JCO from the French
10
group with a total number of more than 1000
11
patients in which overall there has been no
12
combination that has shown a statistically
13
significant difference in overall response rate, in
14
complete response rate, in durable response rate or
15 in
progression-free survival.
16
[Slide]
17
I appeared last before this committee in
18
1999 in relationship to metastatic melanoma. In
19
that setting, it was to introduce the application
20 for
temozolomide. This is an oral equivalent
of
21
dacarbazine that I think no one questions was
22
equivalent to dacarbazine. The
committee did not
34
1 vote
to approve that agent which achieved
2
equivalency in a trial that had been targeted upon
3
superiority. But since that time
I think it has to
4 be
admitted that temozolomide has been the most
5
widely used drug in the community across the
6
country. The FDA briefing that
you have before you
7
suggests that Genasense is, in fact, comparable to
8
temozolomide. I would argue that
it is not.
9
The overall response rate for the
10
temozolomide application was not significantly
11
different. The complete
responses, identical; the
12
durable responses, not detailed; and the
13
differences in progression-free survival with an
14
asymmetrical interval of assessment for the two
15
arms, as Dr. Pazdur has just spoken about,
16
significant but 11 days.
17
The other major difference about
18
temozolomide is that this agent was already going
19 to
be available to the community at large for trial
20
exploration, and the agent that we are going to
21
discuss today will not be available if it is not
22
approved today.
35
1
[Slide]
2
In summary, despite more than 25 years of
3
work and low response rates with the single agent
4
dacarbazine, this agent remains the reference
5
standard for the field. No single
cytotoxic drug
6 nor
any biological agent or combination has been
7
shown to be superior to single agent dacarbazine in
8
relation to survival.
9
Relative to dacarbazine, no large
10
randomized, multicenter comparative study has ever
11
shown a statistically significant benefit in
12
overall response rate, in complete response rate or
13 in
progression-free.
14
High-dose IL-2 is a useful agent that many
15 of
us use for selected patients who lack
16
significant co-morbidity and who are willing to
17
accept its side effects. This
drug is not suitable
18 for
the majority of patients who present to us with
19
metastatic melanoma and is particularly unsuited
20 for
patients who are elderly.
21
[Slide]
22
I would conclude that metastatic melanoma,
36
1
upon which I have focused the last 33 years of my
2
work, is a drug-refractory neoplasm.
We need new
3
agents desperately. Thank you.
4 Study GM301
5
DR. ITRI: Thank you, Dr.
Kirkwood.
6
[Slide]
7
Genasense is an example of a new class of
8
drugs called antisense. Antisense
is fundamentally
9 a
protein knockout strategy. Genasense
inhibits
10
Bcl-2 production. Bcl-2 is a
protein and is
11
believed to be an important mediator of cancer cell
12
resistance to chemotherapy.
Genasense is
13
administered for 5 days before chemotherapy,
14
reduces Bcl-2 production and renders the cancer
15
cell more susceptible to chemotherapy.
In this
16
way, Genasense is postulated to enhance the
17
efficacy of chemotherapy.
18
[Slide]
19
Bcl-2 is ubiquitously expressed by
20
melanoma cells. Five days of
continuous IV therapy
21
with Genasense prior to the administration of DTIC
22
resulted in approximately 70 percent reduction in
37
1
Bcl-2 levels in melanoma cells taken from patients
2
before and after Genasense treatment.
These
3
results provided the rationale for a Phase 3 study
4 in
patients with advanced malignant melanoma.
5
[Slide]
6
This study is the largest randomized trial
7
ever conducted in patients with advanced malignant
8
melanoma. It was an open-label,
multicenter trial
9
involving 139 investigational sites in 9 countries
10
around the world.
11
The primary endpoint was overall survival
12 and
the secondary endpoints included
13
progression-free survival, antitumor responses
14
using computer calculated RECIST based on
15
evaluations of site tumor measurements; durable
16
responses which were defined as responses lasting
17
longer than 6 months; and, of course, safety in all
18
patients.
19
[Slide]
20
Patients received either DTIC at the
21
standard dose of 1000 mg/m
2 or the same dose of
22
DTIC preceded by a 5-day continuous infusion of
38
1
Genasense at a dose of 7 mg/kg/day.
Patients were
2
stratified according to the three major prognostic
3
factors for melanoma, ECOG performance status 0 or
4
1-2; the presence or absence of liver metastases;
5 and
normal or elevated LDH levels. Patients
could
6
receive up to 8 cycles during a treatment phase
7
which were administered every 21 days.
Restarting
8
evaluations were performed at the end of every two
9
cycles.
10
It is important to note that the timing of
11
interval measurements were fixed and similar in
12
both arms, and they were prospectively defined with
13 FDA
agreement, with the temozolomide review issues
14
clearly in mind. Crossover was
not permitted from
15 the
DTIC arm into the Genasense arm, and follow-up
16 was
continued for 2 years in both arms of the
17
study. Patients on the Genasense
arm only could
18
receive up to an additional 8 cycles of the
19
combination therapy in extension protocol GM214 if
20
they achieved at least stable disease by the end of
21 the
treatment phase and it was considered to be in
22 the
best interest of the patient, in consultation
39
1
with the treating physician.
2
[Slide]
3 The statistical assumptions for this
study
4
were based on an overall median survival for DTIC
5 of
6 months which was derived from published
6
reviews. Genasense was postulated
to add an
7
additional 2 months, for total a median survival of
8 8
months; 750 patients would provide 90 percent
9
power to see a difference between groups, with an
10
alpha level of 0.05. It was
assumed that accrual
11
would be constant at 30 patients per month. In
12
agreement with FDA, an analysis was planned when at
13
least 508 deaths had occurred on the study.
14
[Slide]
15
The two groups were balanced for age and
16
gender. The median age of
patients in this study
17 was
60 years but patients ranged in age from 16 to
18
93. Approximately 40 percent of
our patients in
19
this study were greater than 65 years of age and,
20
remarkably, more than 10 percent were more than 75
21
years of age.
22
[Slide]
40
1
The two groups were equally balanced with
2
regard to baseline performance status and
3
approximately half of all patients were symptomatic
4 at
baseline.
5
[Slide]
6
Similarly, the two groups were balanced
7
with respect to the major prognostic indicators
8
including time from initial diagnosis, LDH/disease
9
site distribution and prior immunotherapy which
10 consisted
primarily of alpha interferon
11
administered as an adjuvant therapy in both groups.
12
[Slide]
13
Forty patients who were randomized into
14 the
study did not receive treatment. The
primary
15
reason for this is that in the DTIC arm some
16
patients, later being randomized to the standard of
17
care, were unwilling to travel or withdrew consent
18
once they learned they would not be receiving
19
experimental therapy. The amount
of DTIC delivered
20 to
both groups was equivalent. Overall, the
21
addition of Genasense did not require dose
22
reduction of DTIC.
41
1
[Slide]
2
This is a summary of the efficacy
3
parameters which, taken together, provide evidence
4 for
the benefit of combining Genasense with DTIC.
5 I
will discuss each of these in more detail in
6
following slides.
7
Although not statistically significant,
8
improvement in overall survival was noted for the
9
Genasense group. Statistically
significant
10
improvement was noted in both progression-free
11
survival and response rates, and I will shortly be
12
showing you some interesting updated results
13
regarding complete responses in this study. We
14
also saw a positive trend in patients with durable
15
responses.
16
[Slide]
17
The FDA has raised a number of
18
considerations for the committee's review. These
19
include response rate concordance; the impact of
20
interval assessments on progression-free survival;
21 the
impact of missing data on progression-free
22
survival; baseline differences in prognostic
42
1
factors; and the influence of non-U.S. sites on
2
response rate. I will address
each of these issues
3
separately in the appropriate sections of my
4
presentation.
5
[Slide]
6
This Kaplan-Meier plot of overall survival
7
shows that both arms outperformed expectations.
8
DTIC was associated with a 7.9 month median
9
survival as opposed to the expected 6 months, and
10
Genasense treatment resulted in a 9.1 month median
11
survival. These differences were
not statistically
12
significant. Please note that the
overall survival
13
curves begin to separate at 6 months and the median
14
follow-up at the time of database lock was 7
15
months.
16
[Slide]
17
The addition of Genasense was associated
18
with an overall response rate of 11.7 percent as
19
compared to 6.8 percent for DTIC alone.
This
20
difference is significant, with a p value of 0.019.
21 Use
of the stringent RECIST measurement system has
22
historically reduced response rates in other
43
1
studies by 25-50 percent when compared to
2
investigator determinations.
3
[Slide]
4
It is appropriate at this point to discuss
5 how
responses were calculated in this study.
The
6
investigators did not determine response.
7
Investigators measured lesions and entered these
8
data onto an electronic case report form. The
9
computer then calculated whether the response met
10
criteria for RECIST. RadPharm was
only contracted
11 to
review responding patients. The sponsor
was
12
provided with measurements of target lesions and
13
evaluations of non-target lesions by RadPharm.
14
These measurements were also assessed by the same
15
computer algorithm using RECIST criteria. RadPharm
16
reviewers were blinded as to the treatment arm and
17 all
clinical information in which tumors had been
18
selected by the sites as target lesions.
All marks
19
made by the sites on x-rays were removed.
20
There are three major reasons why RadPharm
21
readings might not have been strictly concordant
22
with the site measurements. These
include the
44
1
evaluation of different target lesions with
2
different measurements, the absence of important
3
clinical information regarding preexisting lesions
4 and
controversy regarding the reporting of normal
5 or
residual lymph node tissue.
6
[Slide]
7
The patient on this slide had extensive
8
liver metastasis at baseline which resolved
9
completely during treatment. This
patient has
10
remained in complete clinical
remission for
11
approximately three years.
12
[Slide]
13
Due to the presence of a persisting liver
14
lesion in the same patient, RadPharm was unable to
15
confirm a complete response. By
procedure,
16
RadPharm was unaware that this was a documented
17
preexisting cystic lesion that was benign. This
18 patient is being cared for by Dr. Hersey who
is
19
here with us today and can answer any questions you
20
might have regarding her treatment course.
21
[Slide]
22
In the next case, which demonstrates how
45
1 the
absence of medical history can confound
2
concordance, a biopsy-proven metastatic lesion of
3 the
frontal sinus was read by RadPharm as
4
incidental sinusitis. Because
this patient had
5
undergone a Caldwell Luck enterotomy with removal
6 of
the inferior turbinate due to metastatic
7
melanoma, RadPharm reasonably assumed that this was
8 an
infectious process and did not confirm the
9
response.
10 [Slide]
11
Because RECIST criteria do not provide
12
guidance for the interpretation of normal lymph
13
nodal architecture at the site of previous disease,
14
RadPharm could not confirm complete response in the
15 next
case and several others like it. Despite
16
complete regression of the tumor next to the blood
17
vessel, here, RadPharm could only assign partial
18
response due to the presence of small residua.
19
The PET scan results for this same patient
20
confirmed complete clinical response and shows no
21
residual evidence of a viable signal post
22
treatment. The FDA did not review
any of these
46
1 x-rays
and based their concordance judgments solely
2 on
raw measurements in percent reductions provided
3 by
the sponsor at their request. I urge the
4
committee to address questions regarding
5
radiographic reviews to Dr. Robert Ford, who is
6
here with us today as an expert consultant in
7
radiology and who personally reviewed all of these
8
films.
9
[Slide]
10
Seventy-one responding patients were
11
evaluated by RadPharm and 60 of these were
12
considered to be evaluable; 11 patients were not
13
evaluable due to the poor quality of photographs or
14
films or the absence of lesions which could be
15
considered measurable by RadPharm.
Five of these
16
cases occurred in the Genasense arm and 6 occurred
17 in
the DTIC arm.
18
Point-to-point concordance for two time
19
point evaluations were available for 38 patients
20 and
give the concordant rate of 63 percent which is
21
consistent with literature citations for
22
evaluations of this nature. Two
additional
47
1
responding patients were confirmed to be responses
2 but
were assessed differently by the site and by
3
RadPharm. Eight cases were
consistent at a single
4
evaluation and were within 10 percent of response
5 at
the second evaluation. Four patients,
such as
6 the
ones I have previously described to you, were
7
easily explained by the absence of appropriate
8
medical history. If we include
only the 40
9
responders confirmed by RadPharm and agreed to by
10 the
FDA on treatment comparison, Genasense is
11
completely consistent to DTIC as demonstrated by
12
odds ratios. If only those 40
responses considered
13 to
be confirmed by both RadPharm and the FDA are
14
included, odds ratios reveal a 91 percent
15
improvement in response rate by RadPharm compared
16 to
an 82 Percent improvement in response for
17
Genasense as reported in the NDA.
18
[Slide]
19
These cases were randomly selected by FDA
20 and
included 40 cases in each arm of the study.
21
X-rays were collected from around the world and
22
included assessments which occurred in the
48
1
follow-up period after NDA cutoff.
As a
2
consequence of this unplanned review of cases,
3
RadPharm was able to identify additional responses
4
which occurred in the follow-up period after NDA
5
cutoff. These important clinical
findings prompted
6
Genta to evaluate all patients in follow-up who met
7
RECIST criteria for response during at least one
8
time point during the treatment phase and all
9
patients who ended the treatment phase without
10
disease progression and who had received no
11
intervening therapy.
12
[Slide]
13
As with response, we observed good
14 concordance
regarding the conclusions about time to
15
progression between the investigational site
16
assessments and RadPharm determinations.
When the
17
site assessments and RadPharm determinations for
18
time to progression are compared, both showed a
19
benefit for the Genasense group.
RadPharm
20
assessments of time to progression in the Genasense
21
group were generally longer than the site
22
assessments.
49
1
[Slide]
2
Six additional responses have been
3
identified which occurred in the follow-up period
4
after the NDA submission and all were in the
5
Genasense group. Only complete
responses are
6
reported since they are the ones most unequivocally
7
associated with clinical benefit and constitute a
8
result not commonly observed with single-agent
9
DTIC. Three of these complete
responses were
10
upgraded from the partial response category and 3
11
were patients with long-standing stable disease.
12
Information regarding these additional responding
13
patients was submitted to the FDA on April 9th of
14
this year.
15
It is important to note that the submitted
16
database has not been updated or altered in any
17
way, nor are we attempting to change the data
18
provided in our NDA. We wish
simply to inform you
19 of
important and frankly unanticipated clinical
20
findings. These responses all
occurred in the
21
absence of other intervening therapies and have
22
been documented by duplicate CT scans using the
50
1
same RECIST criteria as specified in the protocol.
2 The
physicians caring for several of these patients
3 are
here with us today and are able to answer any
4
questions you may have directly.
5
[Slide]
6
Complete responses were evenly distributed
7 by
gender and generally exhibited the same
8
demographic pattern as the overall population.
9
Importantly, one-third of the responses occurred in
10
patients with elevated LDH and half were observed
11 in
the worst AJCC prognostic categories, M1b and
12
M1c.
13
[Slide]
14
Survival for the complete responders
15
ranges from 15 months to more than 3 years on the
16
Genasense arm, and 19 to 21 months on the DTIC arm.
17 The
plus signs denote ongoing responses. Two
18
patients have died, one on each arm of the study.
19
[Slide]
20
The evolution of the complete responders
21 on
this study is shown in this slide. The
two
22
responding DTIC patients are shown in yellow for
51
1
comparison. The solid bar denotes
the database
2
cutoff of August 1, 2003 and is the information
3
contained in the NDA. The dotted
line denotes the
4 date
of the FDA inquiry that precipitated review in
5 the
follow-up period after database cutoff.
6
As you can see, partial responses tend to
7
occur later in the Genasense arm and evolved over
8
time into complete responses. Three
of the
9
Genasense responses, similar to what has been
10
described for IL-2, have been surgically
11
maintained. Once again, all
responses were based
12 on
strict RECIST criteria with duplicate
13
measurements and no patient received intervening
14
therapy.
15
[Slide]
16
Returning now to the data previously
17
reported in the NDA database, the duration of
18
response is presented using a box-and-whisker plot
19 on
this slide. The red line denotes the
median.
20 The
top of the box is the boundary of the third
21
quartile and the bottom is the boundary of the
22
first quartile. As you can see,
the medians are
52
1
similar but an important difference is observed in
2 the
third quartile, resulting in a longer mean
3
duration of response in patients who received
4
Genasense.
5
[Slide]
6
Durable responses, defined as responses
7
lasting at least 6 months, were more than doubled
8 in
the Genasense group, as shown in this slide.
9
[Slide]
10
Median progression-free survival for the
11
Genasense group was 74 days as compared to 49 days
12 for
the DTIC group. The relative risk of
having
13
progressive disease or death was reduced by
14
approximately 27 percent in the Genasense arm.
15
These differences are highly significant, with a p
16
value of 0.0003.
17
Time to progression was performed as a
18
sensitivity analysis for progression-free survival.
19 The
results were very similar and showed
20
approximately a 27 percent reduction in the risk of
21
progressive disease. In this analysis,
11 patients
22 who
died without documented disease progression
53
1
were censored to the day of last lesion
2
measurement. These 11 patients
constitute the only
3
difference between progression-free survival and
4
time to progression in this study, and explain why
5 the
two curves are so similar.
6
[Slide]
7
Genta conducted multiple sensitivity
8
analyses to address possible biases in the
9
calculation of progression-free survival. In all
10
instances the hazard ratios remained stable and all
11
were statistically significant, attesting to the
12
robustness of the observation.
The most common
13 concerns
regarding progression-free survival
14
analyses include the impact of scheduled assessment
15 and
missing data which can potentially be a source
16 of
bias. Several of the methods used by
Genta
17
address these issues and all confirm the conclusion
18
derived from the original planned analysis.
19
[Slide]
20
FDA has performed four analyses using
21
interval censoring techniques.
Hazard ratios are
22 not
reported for this method. Approach number
one
54
1
specifically addresses the issue of assessment
2
schedule bias and remains statistically significant
3 in
favor of Genasense. Approaches two,
three and
4 four
address both assessment schedule and missing
5
data biases taken together.
Approaches two and
6
three remain statistically significant in favor of
7
Genasense. Only approach four,
which represents a
8
rather extreme case assumption, and I will show you
9 an
example of this on the next slide, resulted in
10 an
insignificant p value and would have resulted in
11 the
deletion of almost half of the data.
12
[Slide]
13
Using this example of patient data by
14
interval censoring technique number four all of the
15
data in yellow would have been thrown out because
16 the
investigator failed to repeatedly record the
17
absence of brain metastases. I
would encourage
18
committee members to address any questions you
19
might have for the sponsor regarding this analysis
20
technique to Dr. Janet Wittes.
21
[Slide]
22
In order to address FDA concerns about
55
1
potential differences for baseline variables to
2
affect efficacy endpoints, progression-free
3
survival results and response rates were adjusted
4 for
the variables of age, gender and AJCC LDH
5
disease site criteria. Results
show that both
6
hazard ratios and odds ratios remain stable and all
7
results remain statistically significant. Thus,
8
there was no apparent impact of potential baseline
9
imbalances on results.
10
[Slide]
11
An additional concern has been raised
12
regarding benefit for patients in the United States
13
when response rates are examined by country. This
14
tree plot shows that confidence limits overlap and
15
point estimates are similar for the United States
16 and
non-United States. There is, of course,
17
expected variability in some countries with small
18
sample sizes but no evidence exists that the
19
beneficial effect of the Genasense combination is
20
different in the United States than it is outside
21 the
United States.
22
[Slide]
56
1
In summary, we have demonstrated
2
radiographic concordance and superiority of
3
Genasense regardless of who reviews the x-rays.
4
Progression-free survival was not biased by missing
5
data or interval assessment irregularities. No
6
effect on endpoints was observed related to
7 baseline
demographic variables and similar benefit
8 was
observed for both U.S. and non-U.S. patients on
9 the
study.
10
[Slide]
11
Turning now to safety, adverse events were
12
generally increased in the Genasense arm, as can be
13
expected with add-on therapy. The
committee is
14
referred to the briefing document provided by the
15
sponsor for details of adverse events.
16
Importantly, no new or unexpected adverse events
17
were observed in the study which have not been seen
18
with DTIC alone. We did see an
increase in the
19
incidence of fever, which is a well-known effect
20
related to Genasense as a single agent, as well as
21 an
increase in neutropenia, thrombocytopenia and
22
catheter-related complications.
Safety data were
57
1
regularly and carefully monitored by an independent
2
drug safety monitoring board who at no point
3
identified any safety concerns in the study.
4
[Slide]
5
There is an increased incidence of grade
6
3-4, as well as serious events of thrombocytopenia
7 in
the Genasense arm. The word
"serious" in this
8
context is defined in its regulatory context and
9
generally means the need for hospitalization or the
10
prolongation of hospitalization.
However,
11
bleeding, which is the major clinical consequence
12 of
this laboratory abnormality with grade 3-4
13
bleeding, serious bleeding--serious bleeding
14
related to thrombocytopenia, shows no difference
15
between the arms. Similarly, the
number of
16
patients who required platelet transfusions with
17 the
absolute number of units transfused were no
18
different between the two treatment arms.
19
[Slide]
20
Neutropenia exhibited a similar pattern as
21
thrombocytopenia. The incidence
of grade 3-4 and
22
serious events was increased in the Genasense arm.
58
1
Although higher in the Genasense arm and largely
2
related to the presence of a central line, the
3
incidence of grade 3-4 and serious neutropenic
4
infections was generally low in both groups.
5
[Slide]
6
Not surprisingly, catheter-related
7
complications occurred almost solely in the
8
Genasense arm and the incidence was consistent to
9
that reported in the literature for central venous
10
catheters. Injection site
infections occurred in
11
approximately 4 percent of patients and thrombotic
12
events occurred in approximately 2 percent of
13
patients receiving Genasense, whereas injection
14
site reactions occurred only in the DTIC group
15
where peripheral lines are generally used for DTIC
16
administration. Two patients in
the Genasense arm
17
received their 5-day Genasense dose in 5 hours due
18 to
a mis-programming of the pump. Both of
these
19
patients experienced nausea, fever and
20
thrombocytopenia. Both patients
recovered
21
completely within 48 hours and had no sequelae
22
related to the overdose. Both
patients went on to
59
1
receive the additional cycles of therapy and one of
2
these patients has achieved a PR after 7 additional
3
cycles of treatment. We are
hopeful that
4
subcutaneous and other alternative dosing methods
5 in
development will mitigate the need for a central
6
line and its attendant complications.
7
[Slide]
8
Adverse events leading to discontinuation
9
were increased in the Genasense arm.
However, the
10
majority of events in both arms were related to
11
disease progression. In this
study disease
12
progression could be reported as an adverse event.
13
Importantly, adverse events resulting in death and
14
deaths which occurred within 30 days of the last
15
dose of study drug were no different between the
16 two
treatment arms.
17
[Slide]
18
In summary, this study was the largest
19
randomized trial ever completed in patients with
20
advanced malignant melanoma. The
study was
21
carefully conducted; showed internally consistent
22
results; and demonstrated compelling clinical
60
1
benefit.
2
We believe that we have addressed all of
3 the
study questions given to ODAC for
4
consideration. Finally, we
believe that the study
5
shows consistent clinical benefit, which will be
6
summarized by Dr. Frank Haluska in his closing
7
remarks.
8
In closing, I would like to thank the
9
patients and their families, the physicians, the
10
nurses and the site coordinators who made the study
11
possible. I would also like to
thank the dedicated
12 and
professional employees of Genta who worked
13
tirelessly to contribute to the treatment of cancer
14
patients. Thank you for your
attention. Dr.
15
Haluska?
16 Clinical Benefit Summary
17
DR. HALUSKA: Thank you, Dr. Itri.
18
[Slide]
19
My task today is to provide you with a
20
summary of the data that you have just seen, that I
21
think have been so clearly presented, as well as an
22
overview and some context for the clinical trial.
61
1
[Slide]
2
I think the best way to do this is to in
3 our
minds assume the role of ODAC and if I were a
4
member of ODAC right now I would have two major
5
questions. The first of these is
that the sponsor
6
here has failed to meet the primary endpoint of the
7
study, which is survival--can I still approve this
8
drug? I think the answer to that
question is an
9 emphatic
yes. Dr. Pazdur has already commented
10
that although meeting a survival endpoint is
11
desirable and is the gold standard, the failure to
12 do
so does not preclude approval, and I think that
13 is
germane here.
14
I addition, I think it is important to
15
consider the recent regulatory history of the
16
melanoma field, specifically with regard to IL-2
17 and
temozolomide. IL-2, as you know, was
approved
18
several years ago based on the rate, the quality
19 and
the duration of the responses, data that we are
20
presenting here, and I think these data are
21
stronger because they are the result of a
22
randomized, prospective trial, albeit with
62
1
secondary endpoints.
2
The other drug that I think is relevant is
3
temozolomide and, as Dr. Kirkwood has already
4
explained, the data are better for Genta than for
5 the
temozolomide submission as well. So, I
think
6
that this drug is approvable despite the failure to
7
meet the primary endpoint.
8
The second question that must be on your
9
mind is do the secondary endpoints confer or
10
support the conferral of clinical benefit? Are
11
they strong enough to support approval of this
12
drug? I do think that significant
clinical benefit
13 is
strongly suggested by these data. So,
let's
14
consider that.
15
[Slide]
16
These are I think the most
important
17
endpoints of this study. Again, I
want to stress
18
that they were prospectively identified as opposed
19 to,
for instance, IL-2s which were the result of
20
Phase 2 data.
21
The first of them is the overall response
22
rate. The overall response rate
approaches 12
63
1
percent versus 6.8 percent in the DTIC arm. This
2 is
an improvement. In this field, no
improvement
3
with statistical significance has ever been
4
demonstrated in response rate for advanced
5
melanoma.
6
We have demonstrated improvement in
7
complete responses, 11 versus 2.
This is
8
significant as well and, again, this has not been
9
demonstrated in a reaction study.
I think the IL-2
10
experience is relevant to both of these.
As I
11
said, IL-2 was approved on the basis of the rate,
12 the
quality and the duration of survival. We
have,
13 in
this trial, 9 patients that are alive, an
14
increment that is not seen in the DTIC trial, and I
15
want to point out that IL-2 was approved on the
16
basis of 10. So, this is
certainly in keeping with
17
previous decisions that have been made.
18
The final issue is progression-free
19
survival, 74 versus 49 days, nearly an additional
20
month for patients who are presenting to their
21
oncologist. That is an extra
visit a patient can
22 come
to their oncologist without having been told
64
1
that their disease is progressing.
This, to my
2
mind, is clinical benefit.
3
[Slide]
4
What is the context of these findings?
5
These are the data from the five largest randomized
6
trials that have been conducted in melanoma and the
7
trial in front of you today is the largest. There
8 are
2019 patients that have been treated on these
9
trials and until today there has never been a
10
significant clinical improvement for any of the
11
measures that we are discussing today.
Response
12
rate has not been shown to be improved and it is
13
shown to be improved here.
Complete responses have
14
never been documented in a randomized study to be
15
improved and they are improved here.
And,
16
progression-free survival has never been shown to
17 be
improved and it is improved here. I
think this
18
trial sets itself apart from the progress in the
19
field in the last few years and I think that is why
20 it
requires your careful consideration today.
21
[Slide]
22
To summarize that, patients value
65
1
responses and value complete responses.
The FDA in
2 the
past has made it clear that these are important
3
criteria to consider and, in fact, there are no
4
melanoma drugs approved that have been approved on
5 any
other criteria.
6
You might ask is a 10 percent response
7
rate, or the order of magnitude of 10 percent,
8
important to patients and I think it is with, I
9
think, the recent approval history and data on
10
responses in other malignancies, particularly in
11
lung cancer. The IRESSA
experience that has
12
recently been clarified with data published last
13
week suggests that a 10 percent response rate is
14
clinically important. We
understand the biological
15
basis of some of these responses and a 10 percent
16
response rate can certainly change the field; it
17 can
certainly change a patient's life. So, I
do
18 not
think that a 10 percent response rate in and of
19
itself argues against approval.
20
What about the magnitude of time to
21
progression? A month, I think, is
important. Data
22
that Carey Kilbridge and my colleagues have
66
1
examined with regard to how melanoma patients view
2
their experience strongly suggest that any
3
additional time without being told their disease is
4
progressing or without the presence of disease is
5
important to them. In my opinion,
what the
6
sponsors have shown today constitutes clinical
7
benefit for the melanoma patient.
8
[Slide]
9
What about safety? When we
research a
10
treatment for our patients we do it based on an
11
evaluation of risk versus benefit.
What are the
12
risks of this therapy? The
sponsor has shown that
13
there are no new or unexpected adverse events
14
concomitant to treatment with DTIC and Genasense.
15
There is no difference in the treatment-related
16
deaths between the two arms.
There is an increase
17 in
fever, neutropenia and thrombocytopenia.
Some
18 of
this is likely due to catheter-related
19
complications and this is certainly not the only
20
agent on the market or potentially on the market
21
that would be administered with a pump.
22
Finally, Genasense is still better
67
1
tolerated than other alternatives for melanoma
2
patients and, again, I think a review of the
3
literature is germane here.
4
[Slide]
5
These are three of the trials for which we
6
have good safety data in comparison to the trial in
7
front of you today. They
demonstrate that the rate
8 of
complications for the DTIC arm is certainly
9
similar to what was seen in other studies with
10
regard to grade 3 or 4 neutropenia and grade 3 and
11 4
thrombocytopenia, and certainly the rates of
12
complications that can be attributed to the
13
combination of Genasense and DTIC are less than
14
what we see with other alternatives for melanoma
15
patients. I think that argues
that this is a safe
16
combination and the risk-benefit analysis is
17
completely reasonable to be attributed to therapy.
18
[Slide]
19
Conclusions--I think this is a novel drug.
20 It
is the first of a class of agents that has been
21
shown to be efficacious by several measures. It
22
takes into account our genetic understanding of
68
1
this disease. It is in keeping
with the movement
2 in
the field broadly for targeted therapy and I
3
think that should be taken into consideration.
4
It confers a clinical benefit with DTIC by
5
multiple measures that I think have been reliably
6
demonstrated in this large clinical trial that
7
include response rate, complete responses and
8
progression-free survival. And,
it has a
9
predictable and manageable safety profile.
10
[Slide]
11
Melanoma is refractory to current
12
front-line therapy. You have
heard and I think you
13
will hear further today that we need new agents.
14
This product is safe; it is effective when combined
15
with DTIC to treat stage 4 melanoma.
In other
16
words, this drug works. I think
it is up to you to
17
define today what "works" means but I don't think
18 we
can discard the randomized trial demonstrated
19
improvement in response rate, in progression-free
20
survival and in complete response rate.
21
A final comment--I am supposed to be here
22 as
a dispassionate expert, scientifically objective
69
1 and
clinically removed but I don't think I can
2
completely play that role because I do take care of
3
melanoma patients. The melanoma
field has been
4
criticized for trying to consistently hit the
5
clinical home run. But this
represents progress.
6 It
is incremental progress. It is not a
clinical
7
home run but it is incremental progress, and if we
8 are
ultimately going to make real progress in this
9
disease to cure it, it will require the
10
accumulation of incremental progress.
Allow us to
11
make incremental progress; make this drug available
12 to
our patients. Thank you.
13
DR. PRZEPIORKA: We are going to
hold
14
questions for the first presentation until the FDA
15
presentation has been completed.
Dr. Kane, if you
16
could begin? Thank you.
17
FDA Presentation
18 Medical Review
19
DR. KANE: Thank you.
20
[Slide]
21
Good morning. My name is Robert
Kane. I
22 am
the medical reviewer for this NDA and I will be
70
1
presenting the FDA review along with Dr. Peiling
2
Yang, our statistical reviewer.
3
[Slide]
4
I would like to recognize our primary
5 review
team members for this NDA.
6
[Slide]
7
Randomized, controlled trials
8
prospectively designed with clear, quantitative
9
endpoints statistically analyzed provide the basis
10 to
assess the merits of new drugs. Clinical
11
judgment translates these findings for best patient
12
care. Our presentation today will
include
13
requirements for new drug approval based on federal
14 law
and regulations; aspects of ODAC review of
15
temozolomide which are relevant to today; the FDA
16
examination of the Genasense, oblimersen, NDA; and
17
concluding remarks.
18
[Slide]
19
In the FD&C Act of 1962 substantial
20
evidence of effectiveness was required by Congress.
21
This was defined as evidence from adequate and
22
well-controlled investigations, generally
71
1
understood to mean at least two such studies for
2 new
drug approval.
3
[Slide]
4
The FDAMA legislation in 1997 indicated
5
that one trial may suffice for approval with
6
confirmatory evidence. The
guidance document on
7
effectiveness in 1998 indicated that for a single
8 trial
to suffice it should be of excellent design,
9
internally consistent with highly reliable and
10
statistically strong evidence of an important
11
clinical benefit, such as an effect on survival,
12 and
a confirmatory study might be difficult to do
13 for
ethical reasons.
14
[Slide]
15
New drug approval can take two forms.
For
16
regular approval a sponsor needs to show clinical
17
benefit. Accelerated approval
uses a surrogate
18
endpoint reasonably likely to predict clinical
19
benefit and requires subsequent confirmation of the
20
benefit.
21
[Slide]
22
Here are the currently approved drugs for
72
1
metastatic melanoma. In the past
response rate was
2 the
primary basis, as you have seen and as you have
3
already heard, for hydroxyurea and for dacarbazine.
4
Survival times were, and continue to remain, in the
5
range of 5 to 9 months. More
recently,
6
improvements in the quantity or the quality of
7
survival have served as the basis for approval.
8
Also as you have heard, the aldesleukin,
9
interleukin-2, approval was heavily related to the
10 very long complete responders, some in excess
of 5
11
years. Complete responses will be
abbreviated as
12 CRs
on this slide.
13
[Slide]
14
I would like to remind the committee that
15 the
evidence for interferon supported approval for
16 its
adjuvant use although it is often used in the
17
treatment for metastatic disease.
The temozolomide
18
evaluation by ODAC in 1999 is relevant and
19
instructive for today's review.
20
[Slide]
21
This NDA contained one main open-label
22
study, the primary endpoint of which was survival
73
1
time. It was designed to show a
3-month survival
2
benefit for temozolomide alone over DTIC alone.
3
Secondary endpoints were progression-free survival,
4
abbreviated here as PFS, and response rate, RR.
5
[Slide]
6
The results of this study showed no
7
survival benefit for temozolomide over DTIC.
8
Median survivals were 7.7 versus 6.4 months. For
9
progression-free survival the difference was found
10 to
be highly statistically significant with a
11
log-rank p value of 0.002.
However, the median
12 progression-free
survival difference was only 11
13
days. When an ample size is
chosen for a survival
14
endpoint the statistical significance of small
15
differences in early endpoints can appear
16
magnified. Response rates were
not significantly
17
different.
18
[Slide]
19
Temozolomide was not approved.
The study
20
failed to demonstrate the primary endpoint of
21
survival benefit.
Progression-free survival, a
22
secondary endpoint, was of small magnitude at best.
74
1 No
symptomatic benefit was observed and a proposed
2
post hoc 6-month survival analysis was not
3
convincing.
4
[Slide]
5
For Genta's NDA, here are the
important
6
study dates. The Phase 3 protocol
began in July,
7
2000. The data cutoff date was
August 1, 2003, and
8
this represents excellent accrual to the study. On
9
December 8, 2003 the NDA was submitted for FDA
10
review.
11
[Slide]
12
Genta has just presented their trial
13
design. I would like to emphasize
a couple of
14
points. This was a very large,
multicenter,
15
multinational, unblinded study. This was an add-on
16 of
Genasense to DTIC. Prolonged central
venous
17
access is required for the 5-day infusions of
18
Genasense. Genasense may be
abbreviated as G or
19
G3139 on our slides. The protocol
specified an
20
independent review, a blinded group, to assess
21
responders. Also, the ability to
deal with an
22
ambulatory infusion pump was required.
75
1
[Slide]
2
The primary endpoint was survival.
The
3
design was to detect a superiority in survival.
4 The
protocol included seven secondary endpoints,
5
listed here.
6
[Slide]
7
The trial design was to identify a 2-month
8 median
improvement in survival time from 6 months
9
with DTIC alone to 8 months for the addition of
10
Genasense to DTIC. The primary
analysis for the
11
trial was to be the unadjusted log-rank analysis
12 for
the intent-to-treat population.
13
[Slide]
14
The study disposition of patients showed
15
that less than half the patients were still on
16
therapy after the first assessment about day 42.
17
Most patients went off study because of progressive
18
disease; 44 percent remained on study after the
19
first assessment. As I mentioned,
the data cutoff
20
date was August 1 and analysis occurred at 535
21
deaths.
22
[Slide]
76
1
In the primary endpoint analysis, using
2 the
protocol-specified analysis with the
3
intent-to-treat population, no survival benefit was
4
demonstrated by adding Genasense to DTIC treatment
5
versus DTIC alone. These are the
actual survival
6
results. As you have already
seen, the hazard
7
ratio was 0.89 and the log rang p value for the
8
survival difference was 0.18.
9
Dr. Peiling Yang will now provide a more
10 detailed
examination of the progression-free
11
survival.
12 Statistical Review
13
DR. YANG: Thank you, Dr. Kane.
14
[Slide]
15
As seen in Dr. Kane's presentation, the
16
study failed to demonstrate efficacy in the primary
17
endpoint of overall survival at a two-sided alpha
18
level of 0.05. From a statistical
perspective, an
19
efficacy demonstration based on any other endpoint,
20
such as progression-free survival, would only infer
21 a
false-positive error rate. Despite this
concern,
22 the
secondary endpoint, progression-free survival,
77
1 was
evaluated and the important question is
2
regarding progression-free survival.
3
[Slide]
4
We have doubt regarding the applicant's
5
findings and, second, as Dr. Kane will be
6
discussing, there are questions regarding its
7
clinical significance. This will
be summarized in
8
this presentation.
9
[Slide]
10
My review of the progression-free survival
11 is
as follows, review of applicant's analyses and
12
results; then the major FDA concern about
13 assessment
times; then additional FDA concerns.
14
Let's first review the applicant's
15
analysis and results.
Progression-free survival
16 was
defined as time from the data of randomization
17 to
the date of disease progression or death.
The
18
data of disease progression was recorded as the
19
assessment date when disease progression was
20
documented. If the assessment was
on different
21
days, then the latest date among all assessments
22 was
used by this applicant to represent the
78
1
assessment date in that cycle.
2
[Slide]
3
This slide summarizes the applicant's
4
results. The protocol specified as
secondary
5
efficacy analysis or progression-free survival was
6 the
log-rank test with the missing data imputed by
7 the
last observation carried forward method.
The p
8
value based on this approach was very small.
9
However, in a large trial a small p value can be
10
observed even if the treatment effect is small.
11
During the review process FDA requested the
12
applicant to analyze the data using a different
13
approach by censoring patients at the last
14 assessment date when at least 50 percent of
target
15
lesions were measured if the disease had not
16
progressed yet. The p value based
on this approach
17 was
also very small. However, when analyzed
by
18
this approach the observed median progression-free
19
survival in the combination therapy dropped by 13
20
days and in the control arm dropped by only 1 day,
21 as
presented in this table.
22
[Slide]
79
1
An important question is raised while
2
interpreting the results of the analysis of
3
progression-free survival. Is the
applicant's
4
finding a true finding?
5
[Slide]
6
FDA has a major concern in evaluation of
7
progression-free survival, that is, imbalance in
8
observed lesion assessment times between treatment
9
arms. The next few slides address
this concern.
10
[Slide]
11
Lesions were to be measured every 6 weeks
12
during the treatment phase. In
practice, this did
13 not
always occur. Even when they were
assessing
14 the
planned cycles there were still differences in
15
timing between the two arms.
Because this is a
16
very large open-label trial involving two different
17
regimens, one administered on 6 days and the other
18
only 1 day and because the claimed difference was
19
very small, FDA was concerned that the observed
20
differences in progression-free survival might be
21
affected by systematic bias. One
potential bias
22
could be caused by differences in the time of
80
1
lesion assessments.
2
[Slide]
3
We must remember a critical difference
4
between the analysis of survival and of lesion
5
progression. The date of death,
represented by the
6
star, will not change regardless of the evaluation
7
schedule. With progression
measurement, however,
8 the
date we assign for progression is usually the
9
date of a scheduled visit occurring sometime after
10 the
actual progression date. It should not
be
11
surprising that assessing progression at longer
12
intervals leads to a longer time to progression.
13
[Slide]
14
To address this concern FDA summarized the
15
time from the date of randomization to each of the
16
first 3 observed assessments in this pivotal trial.
17
Included in this summary are those assessments
18
which occurred by the time of disease progression
19 or
death and where there was at least one target
20
lesion measurement. The observed
median times from
21
randomization to each of these assessments were
22
obtained for each treatment arm.
They were 48
81
1
versus 43 days to the first assessment; 94 versus
2 87
days to the second assessment; and 137 versus
3 129
days to the third assessment. The p
values for
4 the
log-rank test comparing the entire curves were
5
also obtained for each assessment.
Note that the
6
difference in timing of lesion assessments shows
7
striking statistical significance, with p values of
8 the
same order of magnitude as the claimed
9
difference in progression-free survival.
This
10
finding raises a concern that all or some of the
11
observed progression-free survival difference were
12
caused by this systematic bias in lesion assessment
13
times.
14
[Slide]
15
These are the times to the first
16
assessment curves. Please note
that these are not
17
time to disease progression curves.
The blue curve
18
represents the combination therapy and the red one
19
represents DTIC alone. On the
horizontal axis we
20
have the time from randomization to the first
21
assessment in days. On the
vertical axis we have
22 the
proportion of patients who had the first
82
1
assessment later at a given time.
As seen here,
2 the
blue curve stayed above the red curve all
3
along, suggesting a systematic delay in the first
4
assessment time in the combination treatment arm.
5
[Slide]
6
Similar patterns were observed in the time
7 to
the second assessment curves.
8
[Slide]
9
And to the third assessment curves.
10
[Slide]
11
Imbalance in assessment times may have
12
impact in several ways on the analysis of
13
progression-free survival. The
first impact is
14
that bias may be introduced in estimating
15 progression-free
survival. Second, with a large
16
trial even a small imbalance between treatment arms
17 may
lead to incorrect conclusions.
18
[Slide]
19
This slide illustrates the first impact.
20 A
hypothetical example is given here to illustrate
21 how
imbalance may be introduced in estimating
22
progression-free survival. In
this example,
83
1
suppose that the actual day of disease progression
2 was
day 35 post randomization for both patients,
3 one
in the control arm and the other in the
4
experimental arm. However, the
first assessment
5 for
the patient in the control arm was on day 42
6 and
for the patient in the experimental arm it was
7 on
day 48. The recorded days of
disease-free
8
progression will be on days 42 and 48 respectively.
9
These recorded days, not day 35, will be the
10
observations used in the analysis.
11
[Slide]
12
This slide illustrates the impact of
13
systematic bias by a simulation study.
In the
14
simulation study progression-free survival was
15
generated from identical distribution in both arms
16
with a median of 50 days and 300 subjects in each
17
arm. However, a systematic
increase by 2 days in
18
assessment times in one arm was introduced. In 98
19
percent of the 5000 simulations p values were less
20
than 0.05. This illustrates that
even with a small
21
imbalance in assessment times between two arms the
22
chance of falsely concluding treatment effect can
84
1 be
very high when, in fact, there is no treatment
2 effect
at all, also the chance of incorrectly
3
concluding increases as the sample size increases.
4
[Slide]
5
An additional FDA concern is about missing
6
data. Missing data was observed
in both treatment
7
arms, especially for non-target lesions which also
8 had
an influence on the determination of disease
9
progression. In this study lesion
assessments were
10 not
always performed in planned cycles.
Also,
11
lesions were assessed at baseline or assessed post
12
baseline. In the presence of
missing data bias
13
could be introduced in estimating treatment
14
effects, especially in an open-label study as this
15
is. This is a common problem in
assessing
16
progression in most of the studies.
17
[Slide]
18
This slide summarizes the progression-free
19
survival findings. The claimed
progression-free
20
survival benefit in the combination therapy over
21
DTIC alone may not be a true finding because of
22
imbalance in assessment times between treatment
85
1
arms. The true progression-free
survival benefit
2 of
the combination therapy over DTIC therapy alone
3 was confounded by imbalance in assessment
times
4
between treatment arms. Thus,
true treatment
5
effect with respect to progression-free survival
6
cannot be isolated. The chance of
falsely
7
inferring progression-free survival benefit could
8 be
high. Even if there was, indeed, no
benefit, it
9
will be magnified by increasing the sample size.
10
Missing data is always a concern in oncology
11
studies evaluating progression as an endpoint. The
12
confidence in the amount of difference in
13
progression-free survival is diminished in the
14
presence of missing data and may allow introduction
15 of
bias, especially in an open-label study.
16
[Slide]
17
Finally from a statistical perspective,
18
this large randomized, open-label study failed to
19
demonstrate the protocol specified primary efficacy
20
based on the overall survival benefit with respect
21 to
the secondary efficacy analysis of
22
progression-free survival because of systematic
86
1
bias in ascertainment. It is not
clear whether the
2
benefit of progression-free survival in the
3
combination therapy over DTIC alone exists. If it
4
exists, the magnitude is uncertain.
Also, there
5 are
multiplicity issues with analyses conducted to
6
support the efficacy. Dr. Kane
will address the
7
clinical relevance.
8 Clinical Relevance
9
DR. KANE: Dr. Yang has provided a
10
detailed assessment of some of the concerns related
11 to
progression-free survival.
12
[Slide]
13
To summarize these concerns, assessments
14 in
this study were done at 6-week intervals.
The
15
progression-free survival difference, however, was
16
only in the range of 2-3 weeks.
The
17
progression-free survival difference is highly
18
statistically significant but may be fully
19 accounted
for by asymmetry in the timing of
20
assessments between the two arms.
The magnitude of
21 the
effect size is uncertain. The real
problem is
22
what is the clinical relevance.
87
1
[Slide]
2
The Division examined all of the secondary
3
endpoints of the protocol for the possibility of
4
patient benefit, given the fact that the overall
5
survival analysis failed.
6
[Slide]
7
We will next look at the response rates
8
among the secondary endpoints.
The data submitted
9 at
the time of the original NDA submission and
10
analysis, as has been presented here, indicated
11
that the Genta investigator-determined responses
12
were derived from an algorithm using tumor
13
measurements from the case report forms.
In that
14
examination, 11.7 percent of patients were reported
15 as
responders to the combination versus 6.8 percent
16
with DTIC alone. The p value for
this difference
17 was
0.018 and the actual difference was just under
18 5
percent.
19
The study protocol also called for a
20
blinded independent review and confirmation for all
21 responders.
The protocol stated that all
22
radiographs, as well as photographs of cutaneous
88
1
lesions, were to be provided to this review group.
2 The
blinded independent reviewers, as you have
3
heard, reported different response rates, 6.7
4
percent response for the combination versus 3.6
5
percent for DTIC alone, a difference of 3.1 percent
6 and
of borderline significance. Ordinarily,
7
adjudication by an independent review is considered
8 to
be the definitive response rate.
9
[Slide]
10
Some of this discordance may be due to
11
technical difficulties, such as providing the
12
independent review group with the appropriate
13
images. However, we must point
out that 5 complete
14
responses, which constituted all of the responses
15 in
the initial NDA submission identified by the
16
Genta site investigators--there were 3 in the
17 combination
arm and 2 in the DTIC alone arm. None
18 was
adjudicated as complete responses by the
19
independent review. Forty-four
percent of the
20
responders by the Genta site investigators were
21
determined as not assessable or unconfirmed by the
22
independent review. For 49
percent there was full
89
1
concordance for the response category between Genta
2 and
the independent review.
3
[Slide]
4
You have also heard that on April 9th--a
5
couple of weeks ago--Genta provided new data on
6
responders. This new data is
being examined.
7
There are problems with data that is developed
8
outside of the study protocol.
There can be
9
ascertainment bias between arms when an analysis is
10 not
prospectively planned. Subsequent
therapies,
11
such as surgery not being part of the protocol
12
treatment, may not be applied symmetrically.
13
[Slide]
14
Turning to duration of response, another
15
secondary endpoint, this is Genta's analysis. This
16
data is skewed data and, therefore, we refer to the
17
median to describe it and the medians are quite
18
similar.
19
[Slide]
20
For durable response rate Genta has
21
provided this analysis. This was
a prespecified
22
secondary endpoint. The
difference was not
90
1
significant.
2
[Slide]
3
Performance status is a measure of
4
functional capacity. There were
no differences in
5
performance status observed between study arms to
6
suggest a benefit for adding Genasense to the DTIC.
7
[Slide]
8
For tumor-related symptoms, there were no
9
differences in symptoms observed between study arms
10
during the treatment.
11
[Slide]
12
This slide introduces the adverse events
13
which represent the toxicity safety endpoint for
14 the
study. You have heard from Dr. Itri that
the
15
grade 3-4 adverse events, the serious adverse
16
events, and the adverse events leading to
17
discontinuation all were increased with the
18
addition of Genasense to DTIC.
Since the DTIC
19
doses were the same, the increased toxicity is
20
likely due to the Genasense.
21
[Slide]
22
This represents the hematologic toxicity
91
1
which you have already heard.
There was more grade
2 3-4
neutropenia and thrombocytopenia on the
3
combination arm.
4
[Slide]
5
For non-hematologic toxicity, all adverse
6
events were more frequent on the combination arm
7
with the addition of Genasense.
8
[Slide]
9
In total, there were 18 patients with
10
upper extremity thrombosis on the combination arm
11
compared to 3 on the DTIC alone arm.
12
[Slide]
13
In summary, the Genasense trial failed to
14
achieve its primary protocol-specified endpoint.
15 No
survival benefit was demonstrated with the
16
addition of Genasense to DTIC compared to DTIC
17
alone. The efficacy of the
control arm, DTIC
18
alone, is consistent with that of other studies.
19
[Slide]
20
Looking again at the secondary endpoints,
21
these are usually considered to be exploratory and
22 for
progression-free survival there is no precedent
92
1 for
progression-free survival as evidence of
2
clinical benefit for metastatic melanoma. This may
3 not
be a true finding. The progression-free
4
survival difference between the two arms may be 13
5 or
25 days depending on which censoring technique
6 is
chosen for missing data. The clinical
relevance
7 is
uncertain.
8
[Slide]
9
For response rate, the difference from
10
DTIC alone may be in the range of 3-5 percent. No
11
complete responses in the original NDA submission
12
were confirmed by the independent blinded review
13
committee. The clinical relevance
of this result
14 is
uncertain. Thus far, response rates in
these
15
ranges have not conferred survival benefits for
16
metastatic melanoma. For the
durable response
17
rate, no significant difference.
Response
18
durations were practically identical.
19
[Slide]
20
For performance status no benefit was
21
observed from the addition of Genasense to DTIC
22
over DTIC alone. Symptomatic
benefit was no
93
1
different. There is greater
toxicity with the
2
Genasense combination than for DTIC alone. Thank
3
you.
4 Questions from the Committee
5
DR. PRZEPIORKA: Thank you for the review.
6 We
are now going to open the session for questions
7 to
either the sponsor or to the FDA. Dr.
Cheson?
8
DR. CHESON: I am sure the 11 or
so
9
patients out there still in remission will be
10
disturbed to know that modeling suggests that they
11
shouldn't be there. We have heard
some difficult,
12
complicated analyses of modeling suggesting that
13
what we heard from the elegant presentation from
14 Dr.
Itri and her co-workers might not be as
15
clinically relevant. So, we have
one side
16
suggesting one set of outcomes showing clinical
17
benefit, then the computer modeling and the FDA
18
suggesting perhaps that these are not reliable. I
19 would like to hear from the company, from Dr.
20
Wittes, their side of this spin.
21
DR. WITTES: The issue about the
potential
22 for
bias that can come from interval censoring and
94
1
from missing data we knew about and, in fact,
2
looked at--I need the slide, yes, that is the one.
3
[Slide]
4
In fact, that is why we did some of the
5
sensitivity analyses. These
sensitivity analyses
6
look at three different kinds of things, the
7
missing data and the interval censoring, and the
8
last three are the ones that look at interval
9
censoring, the by-cycle analysis, the assumed
10
progressive disease, back to the scheduled
11
visit--these are three different ways of trying to
12
adjust for the interval censoring.
What you see is
13
some changes in hazard ratio but quite similar to
14
what they were before and then statistically
15 significant
p values.
16
[Slide]
17
Next slide, CC49--the FDA's approach for
18
interval censoring, which is a method due to
19
Michael Fay, is a non-parametric approach. It is a
20
score statistic and, again, the p value remains
21
statistically significant. So,
yes, there
22
certainly is a differential time to measurement in
95
1 the
two groups but analyses that adjust for that
2 time
still show a statistically significant
3
benefit.
4
DR. PRZEPIORKA: Dr. D'Agostino?
5
DR. D'AGOSTINO: Janet, the
procedure the
6 FDA
used is not unreasonable. I am asking a
7
question but it is a set of assumptions that could,
8 in
fact, underlie some of the differences we see,
9 and
I guess the point that the FDA was making, I
10
thought, was that you could chip away at these
11
differences not only in statistical significance
12 but
magnitude of difference, clinical difference,
13 and
that I think should be taken into account with
14 the
interpretation of these techniques.
15
DR. WITTES: I agree, Ralph, but
can we go
16
back to that 49?
17
[Slide]
18
Here is the chipping away. I
mean, the
19
chipping away is to look at both the interval
20
censoring and the missing data. I
think if you
21
approach four, which is the one that is most
22
chipped, if you look at what that does, it is the
96
1
Michael Fay approach to interval censoring plus a
2
very conservative method for missing data, and let
3 me
describe that a little bit because I think it is
4
important to know what happens here.
5
There are basically three kinds of missing
6
data. There are those that Dr.
Itri showed where
7
there is an assessment, it is clear and then you
8
don't keep on looking at that--the no lesion. That
9 is
one source. There is another kind of
missing
10
data where you have an assessment.
At the next
11
assessment you don't measure that lesion and then
12
subsequent to that you do measure it and there is
13 no
progression. So, to me, that isn't
really
14
missing. If you take away those
two and leave the
15
missing data where you really can't know whether
16
there is an assessment or not, this method becomes
17 an
0-3 again. So, I think if you chip it
away you
18
still get evidence of benefit in progression-free
19
survival.
20
The other thing to remember is that from
21 the
point of view of complete responses there is no
22
issue at all about either interval censoring or
97
1
missing data.
2
DR. PRZEPIORKA: Dr. D'Agostino?
3
DR. D'AGOSTINO: But just again
though, we
4 are
left in the dilemma of how do you respond to
5 the
data as collected, as the assessments were made
6 and
so forth, and there is uncertainty in terms of
7 how
comfortable some of us are with the p values.
8 I
think also with a large study you can generate
9 very
large p values with small differences and
10
maybe some of that is here also.
Again, p values
11 are
important but there is clinical significance
12 the
way these numbers draw closer together by, I
13
think, relatively comfortable assumptions that is
14 of
concern I think.
15
DR. WITTES: I think someone else
should
16
address the clinical significance.
17
DR. PRZEPIORKA: Dr. Temple?
18
DR. TEMPLE: Janet, one of the
things
19
about 0.003 is that you don't worry about
20
adjustment for multiplicity and stuff like that.
21 It
kind of blows you away. But with the
smaller p
22
values that you get from some of the other things
98
1 you
did that might become an issue. Do you
have a
2
view as to how one should take into account the
3
fact that this is not the primary endpoint? It is
4 one
of at least several things one could have done.
5 What would you say the right kind of
adjustment
6
would be in a case like that, assuming that some of
7 the
closer to 0.05 p values were the ones that
8
might count?
9
DR. WITTES: Yes, I don't know the
answer
10 to
that. I mean, if the question is what is
the
11
type-1 error of this study, I think one can't
12
really answer that question. Of
course, one looks
13 at
consistency. One worries about the
potential
14 for
bias and, again, I feel that those complete
15
responses kind of avoid--they become a different
16
kind of criterion. But if you ask
me what is the
17
type-1 error rate, I don't know.
18
DR. PRZEPIORKA: Dr. D'Agostino?
19
DR. D'AGOSTINO: Just again, when
you look
20 at
the secondary endpoints after you have a failure
21 in
the primary endpoint, the whole
22
interpretation--just to reinforce what you just
99
1
said, no one around this table is going to be able
2 to
put a real p value on any of these things that
3 we
have given that the primary didn't turn out to
4 be
statistically significant.
5
DR. PRZEPIORKA: Any other
questions from
6 the
committee? Dr. Hwu?
7
DR. HWU: I have a question for
Dr. Itri
8
regarding the design of this trial, especially the
9
regimen used in this large trial for the
10
experimental arm. The initial
scientific
11
indication of this incremental improvement in the
12
treatment of melanoma was based on the Phase 1 and
13 2
trial, which was published in Lancet by Jansen
14 and
colleagues in 2000. The Phase 1 and 2
trial
15
design was extremely careful.
They screened the
16
patients who had shown in tissue increased
17
expression of Bcl-2. Also, the
pharmacokinetic
18
study was done very carefully and was a clinical
19
correlate of the tissues at the level of decrease
20 of
Bcl-2 expression. Also, there is
correlation
21
with responses.
22
The regimen used in that trial was very,
100
1
very reasonable in design. They
were giving
2
infusion on day 1 to day 14, continuous infusion.
3
Clearly by day 5 the Bcl-2 expression was maximally
4
down-regulated. DTIC was given
from day 5 to 9 in
5
divided doses of 200 mg/m
2 every day for 5 days.
6 In
other words, when DTIC is infused in patients,
7 the
G31 and 39 Genasense treatment also continues.
8
Now, the response was clearly shown in the
9 M1a
group, the patient with skin metastases or
10
lymph node metastases. No
response was noted in
11 the
lung or visceral organs. However, the
12
responses were impressive. Even
one patient who
13 had
prior DTIC had a partial response.
14
My question to Dr. Itri is why we changed
15 the
protocol which has clearly demonstrated
16
scientifically that it worked as a target therapy
17 and
now we have changed to 5-day infusion of
18
Genasense followed by 1 infusion of DTIC and even
19
forgot that DTIC is not an active chemotherapy
20
agent by itself; it requires hepatic activation to
21 its
active metabolite MTIC? We do know that
the
22
company provided a pharmacokinetic study that, yes,
101
1 the
continuous infusion of Genasense that achieved
2 the
maximal plateau level within 10 hours if you
3
were giving it at the 7 mg/kg/hour rate--I am
4
sorry, per kilogram--however, once the infusion
5
stopped, less than 10 hours later the level for the
6
Genasense clearly dropped to what we call the
7
biological active level of I think 1 mcg/L.
8
So, I would like to know before we launch
9
this large Phase 3 trial are there any other Phase
10 2
studies, other than the safety, well-tolerated
11
5-day infusion by 1 day of DTIC, that have shown
12
that there is tissue correlation and also efficacy
13 as
shown by the Phase 1 and 2 trial. Thank
you.
14
DR. WALL: I am Dr. Ray Wall, from
Genta.
15 Dr.
Hwu, I think I will take a whack at those
16
questions since I was around at the time the study
17 was
done and took it with Dr. Haluska down to FDA,
18 and
Dr. Itri was not.
19
The Genasense study was informative.
I
20
would point out to the committee it was a Phase 1
21
studies that looked at a couple of different doses
22 of
Genasense at that time and also looked at a
102
1
couple of different routes of administration, both
2
subcutaneous administration as well as continuous
3 IV
infusion. So, it was Phase 1 and it was
a total
4 of
12 patients. It was published in Lancet
in year
5
2000.
6
What we had found both in that study and
7
also in a variety of other studies, some of which
8 are
presented in your briefing book, are a couple
9 of
things with respect to the biological activity
10 of
the drug. The pharmacokinetics are very
well
11
described and I will skip them for the time being.
12
What we see in human tumor cells
13
subsequent to administration of Genasense is that
14 the
onset of the down-regulation of Bcl-2 at the
15
protein level, not the RNA level but of the protein
16
level seems to occur at least as early as day 3 and
17 is
maximal at day 5. The one other thing
that had
18
been a very, very important driver of our clinical
19
schedule is that the continued administration of
20
Genasense beyond day 5, if the dose is not changed
21 you
do not seem to get any further down-regulation
22 of
Bcl-2 at the protein level.
103
1 I didn't bring a lot of blots in
my back
2
pocket here but I think I can show you one from a
3
melanoma patient, if I can have MA-25, please?
4
[Slide]
5
This is a Phase 1 study looking at a very,
6
very low dose. This is a dose
that is about 20
7
percent of our Phase 3 doses, and this is from the
8
Jansen study looking at continuous infusion over a
9
14-day period. Again, you see
maximal
10
down-regulation by about day 5 and, despite the
11
fact that the infusion is continued, you don't see
12 any
further decrease in the down-regulation of
13
Bcl-2 protein effect. These are
human tumor cells,
14
serial biopsies of patients with malignant
15
melanoma.
16 So, from these data and from other
data
17
that have been obtained from a variety of other
18
patients and other cells, both malignant cells as
19
well as normal cells, that molecular information
20 has
been used to drive the clinical studies,
21
including the one that you have seen today.
22
So a couple of things, one is we use
104
1
rather short infusions to maximize the
2
down-regulation of Bcl-2 so that that effect is
3
maximal at the time that chemotherapy is
4
administered and we don't continue beyond. Dr.
5
Tony Tolcher, who actually is in the audience, has
6
done some of the best scheduling work but, again,
7
modeling preclinically, suggesting that when you
8
administer Genasense with chemotherapy the effect
9 is
maximized when you administer Genasense in
10
advance of chemotherapy. The
second thing that he
11 has
shown is that there seems to be no advantage to
12
overlapping Genasense with chemotherapy.
The final
13
observation from the Tolcher lab is that if you
14
reverse the sequence, if you give Genasense after
15
chemotherapy is administered, then you basically
16 eliminate the synergistic effect. So, the
17
constellation of these kinds of pharmacodynamic
18
events have driven the schedules that you have seen
19
here today in Phase 3.
20
DR. PRZEPIORKA: Before you leave
the
21
podium, just one more question to follow-up, how
22
long is the effect once the infusion is
105
1
discontinued?
2
DR. WALL: As was pointed out, the
3
half-life of this drug is around 3-4 hours and
4
fundamentally disappears probably by about 10-12
5
hours. The data are a little
fragmentary and
6
mostly derived from in vitro cell culture studies,
7 but
it does look like the half-life of Bcl-2
8 protein is in the order of 16 to about 22
hours.
9 So,
you would expect that if you get complete
10
shut-down of Bcl-2 production by knocking out the
11
messenger RNA, then pharmacokinetically within 5
12
half-lives or so you should have no protein within
13 the
cell, and recovery would be equally as rapid as
14
soon as it is shut back on.
15
DR. PRZEPIORKA: Dr. Temple?
16
DR. TEMPLE: Dr. Itri or others,
there was
17 a
lot of discussion about the responses.
You
18
clearly had two different ways of calculating
19
responses, one based on investigators and the other
20
based on RadPharm. My presumption
was that the
21
RadPharm analysis existed because the study was
22
open and that is a common thing to do, to have a
106
1
blinded analysis of the response rates.
In your
2
presentation though I gather you were disappointed
3
with what RadPharm produced and you considered it
4
inaccurate. Could you clarify the
intended role,
5
what happened and whether you think there ought to
6 be
a further blinded analysis, or what?
This is a
7
somewhat unusual situation and it wasn't clear what
8 the
original intent was. As Dr. Kane said,
usually
9
when you have a group like that, they are the
10
primary analysis. Was that not
true? Just what
11 was
the arrangement?
12
DR. ITRI: That was not true here.
13 DR. TEMPLE: Then why did you do it?
14
DR. ITRI: The response per
statistical
15
analysis plan was RECIST measurements based on
16
investigational site measurements that were then
17
calculated by computer to see whether or not they
18 met
criteria for a partial response or a complete
19
response. That is primary and
that is what is
20
reported.
21
The use of RadPharm--and I think it is
22
important to note that it was only responding
107
1
patients that they looked at so if we were going to
2
rely on RadPharm to actually give us a response
3
rate for the study they would have had to review
4
everyone. They were really used
by us for quality
5
control purposes. We wanted to
make sure that the
6
relative numbers we were seeing were consistent
7
with what has been reported in the literature; that
8 the
concordance rates weren't really out of whack.
9 I
think that the best person to speak about this is
10 Dr.
Ford because he can put this into real context
11 and
explain what the literature shows, and really
12 how
we stack up in terms of other studies that have
13
utilized a similar review. Is
that okay?
14
DR. TEMPLE: Anything is okay, but
you
15
have two somewhat separate, somewhat different
16
calculations based on the ones that went to them.
17
Usually that is distressing and I guess the further
18
question I have is do you have some way of
19
resolving this? Should this be
subjected to
20
another blinded review where people get the whole
21
files, or something? I mean, as
it is, you can see
22 why
it is sort of troublesome. For example,
all of
108
1 the
complete responses they didn't think were
2
complete responses although you feel that complete
3
responses are very important for the reasons Dr.
4
Cheson mentioned earlier. That is
troublesome, and
5 now
you have found more which we haven't had a
6
chance to review yet, but the same problem could
7
arise there too. So, it does seem
important to
8
figure out what it all means.
9
DR. ITRI: I really think you need
to talk
10 to
Dr. Ford about this.
11
DR. TEMPLE: Whatever you like.
12
DR. ITRI: But the other issue is
that,
13 you
know, if the agency would like us to submit
14
these x-rays for review and if that would make you
15
more comfortable, we would be totally willing to do
16
that. We believe that what is
being called lack of
17
concordance really relates to the fact that Dr.
18
Ford is going to elucidate now.
And, it would not
19 be
a problem; we would be so happy to sit with
20
anyone and give you the clinical data that supports
21
this because these are real and the patients are
22
alive, most importantly. So, we
would welcome a
109
1
chance to sit down and review these x-rays.
2
DR. TEMPLE: While you are at
that, that
3 is
the second question I was going to ask you and
4 maybe
you want to answer them both. The
survival
5
curves don't seem to have different tails on them.
6 So,
I am a little confused about where the
7
long-term survivors you are referring to come from
8 if
they are not in the survival curve, or maybe the
9
curve has been extended.
10
DR. ITRI: We provided update
survival
11
information to the agency--
12
DR. TEMPLE: I just need the one
you
13
showed though.
14
DR. ITRI: Well, that was an early
cutoff
15 so
we don't really know what the tail is doing.
16
That was the 7-month median.
17
DR. TEMPLE: It is really Dr.
Cheson's
18
question I am following up on, if there were a
19
small subset of people that got really important
20
responses, wouldn't you see a difference in where
21 the
tails end up?
22
DR. ITRI: It might be too early
to see it
110
1 on
that curve.
2 DR. TEMPLE: Well, that means they are in
3
both groups then. There are
long-term survivors in
4
both groups. Is that right?
5
DR. ITRI: There are some
long-term
6
survivors.
7
DR. WITTES: It depends on the
nature of
8 the
censoring, where the censoring is. So,
some of
9
that could be showing up before the edge of the
10
tail occurs because they haven't been followed long
11
enough. I mean, the fact that
they come together
12
doesn't eviscerate the point. You
have to look at
13
where the specific events occurred relative to
14
censoring.
15
DR. TEMPLE: That is fair
enough. There
16 was
reference to at least some people who were
17
getting really spectacular benefits and I would
18
have thought that would show up as curves where the
19
flat part is here on one and the flat part is below
20 on
the other.
21
DR. WITTES: They are censored.
22
DR. TEMPLE: They are censored
because
111
1
they haven't been on long enough--
2
DR. WITTES: It is like three
years.
3
DR. FORD: Well, thank you very
much for
4 the
opportunity to address the committee on this
5
topic, the topic at hand being how does an
6
investigator who sees the patient on a daily basis
7 or
a regular basis assess response compared to how
8 an
independent review facility would assess
9
response in the same patient in a remote location,
10 not
having access to the clinical information.
11
I think that there is little written in
12 the
medical literature about this topic, but there
13 are
two particular studies that I would like to
14
review kind of as a background for this discussion.
15 The
first was a study that was published in the
16
Annals of Oncology in 1997. The
author was a
17
radiologist and that was a review of a 100-patient
18
ovarian cancer trial. In that
review there were 24
19
claimed responders who were reviewed by an
20
independent review facility and in that instance
21
there were 14 patients who were concordant, that
22 is,
deemed to be responders by the independent
112
1
review facility and deemed to be concordant with
2 the
investigator.
3
There was a second study that was done,
4
also published in 1997 in the Journal of Clinical
5
Oncology. It was a review of a
renal cell trial
6
where there were 133 subjects who were reviewed.
7 In
that review an independent review facility
8
reviewed those studies and the responses were
9 concordant in 62 out of those reviews. In that
10
article you can see the concordance, that is, site
11
same PR to independent review facility saying PR
12 was
approximately 60 percent, and in the second
13
study it was lower, on the order of 48 percent.
14
Now, with that as a background, there is a
15
significant difference in the methodologies in
16
which those reviews were performed.
That is, in
17
those examples the investigators who enrolled the
18 patients
in the trial were actually part of the
19
review process. A radiologist sat
down with the
20
films, made the measurements and reviewed the
21
images in concert with the physicians who knew much
22
more about that patient, that is, had the
113
1
additional clinical history that the radiologists
2
would have at the time of the review.
3
Now, that as a background, discussing the
4
current study, the current study was a radiology
5
only review. When it was
performed there was no
6
clinical information provided. In
that instance,
7
even in that particular setting the concordance was
8 63
percent. So, 63 percent of the time that
the
9
investigators assessed the response on this trial,
10 the
independent review facility assessed the same
11
response.
12
DR. TEMPLE: When they are
different how
13 do
you know which one is right? When they
are
14
different, non-concordant, how do you decide which
15 one
is right? I am sure I understand that
16
different groups will reach different conclusions.
17
Sometimes these special committees have a
18
tie-breaker when they don't agree. But what is one
19
supposed to do that when they are non-concordant?
20 How
do you decide which is true?
21
DR. FORD: Well, in this
particular
22
setting the investigator-determined response was
114
1
chosen.
2
DR. TEMPLE: When? I mean, was this
3
prospectively defined in the protocol how any
4
discrepancies were going to be handed?
5
DR. ITRI: Yes, it was.
6 DR. TEMPLE: So, the protocol was clear
7
that the investigator-determined conclusion, or the
8
analysis based on the investigator--
9
DR. ITRI: The investigator
measurements
10
were fed into the computer and that is what was to
11 be
used for determination of response.
12
DR. PRZEPIORKA: Dr. Rodriguez?
13
DR. RODRIGUEZ: Yes, this is a
follow-up
14 to
the question by Dr. Hwu because I didn't hear
15 the
response to part of her question, that is, you
16
know, this is a biologically targeted agent and one
17
assumes that one is going to look for the
18
appropriate target or that one would select
19
patients who are appropriate to be treated with
20
this drug. I didn't hear whether
all patients
21
entering on the study were screened, if their
22
tumors were screened for expression of Bcl-2 or if
115
1
there had been an attempt to quantitate category of
2
patients because, obviously, some patients are
3
going to be appropriate for trial and others are
4
not. Was that done?
5
DR. WALL: That is a very good
question.
6 Can
I have slide MA-18, please?
7
[Slide]
8
The challenge with Bcl-2 is the ubiquity
9 of
Bcl-2 expression in melanoma. So, this
is not
10
comparable, for instance, with HER2 expression in
11
breast cancer in which the incidence of expression
12 in
advanced cases is on the order of 20, 25 percent
13 so
that you would not want to treat 100 percent of
14
women. You could theoretically
benefit 25 percent
15 so
the absolute response rate would be 5 percent of
16
your total. In general, we chose
melanoma because
17 of
the very, very high prevalence of expression
18
which in these studies, whether you look at
19
immunohistochemistry, which is the blue bars, or
20
RT-PCR of excised specimen, you are talking about
21 something in the range of 90, 95 percent
expression
22 of
tumors.
116
1
So, the kinds of correlations that you are
2
going to be able to make with respect to
3 over-expression
we thought, going into this study,
4
were going to be extremely limited due to the very
5
high prevalence of baseline expression.
Again, it
6
certainly influenced our choice of melanoma as one
7 of
the early targets for this particular disease.
8
After that it is not clear where you could go if
9 you
were going to look at percentage
10
down-regulation. That meant
serial biopsies of
11
fresh tissues from multiple sites, handled very,
12
very carefully, centrally managed, exponential
13
increases in cost and ability to manage--that
14
simply overwhelmed us as a small company. So, we
15
figured we would pick a big tumor in which would be
16 an
unquestioned level of very, very high expression
17 at
baseline but it did preclude the ability to make
18
subset selections based on--at least at the stage
19 we
were dealing with this in 2000--Bcl-2 expression
20 per
se.
21
DR. PRZEPIORKA: Dr. Hwu?
22
DR. HWU: I agree that choosing melanoma
117
1 as
this malignancy is very important based on what
2 we
know of Bcl-2 over-expression. My
question to
3 you
that you didn't answer is based on your current
4
regimen with some 300 patients.
Have you any data
5 to
show that it clearly reproduced your finding in
6 the
previous Phase 1 and 2 using completely
7
different regimens?
8
DR. WALL: Well, the Phase 1
study, as you
9
know, did not show correlations.
It really was not
10
appropriately powered to look for correlations
11
between baseline Bcl-2 expression and percentage of
12
down-regulation. That is very
difficult to model
13
even preclinically. I am not sure
I am answering
14
your question.
15
DR. HWU: I don't agree that that
is not
16 the
conclusion from the publication. Clearly
the
17 CR
person that has the highest incremental decrease
18 of
Bcl-2 is the percentage of decrease; it is not
19 the
total amount of expression. That is what
I
20
learned from the paper.
21
DR. WALL: I think you need to
keep in
22
mind that it is a Phase 1 study.
That patient got
118
1 a
rather low dose. The majority of
patients were
2
actually not serially sampled.
And, the ability to
3
make inferences with respect to those kinds of
4
correlations with a total N of 12 is I think very
5
problematic.
6
DR. HWU: To make a correction,
the
7
patient got the highest dose level of 6.5 and she
8 had
70 percent--
9
DR. WALL: And that blot was shown
to you,
10 by
the way.
11
DR. HWU: --and the patient had
never
12
received any chemotherapy prior either.
13
[Slide]
14
DR. WALL: Right, and here is the
blot
15
from that patient that Dr. Itri showed.
I think
16 the
major point, however, is with an N of 1 in a
17
sample size of 12 in a Phase 1 study we didn't feel
18
like we could make inferences. I
would say that
19 one
of the advantages of being an oncologist is
20
that you can fall back on issues related to
21
maximally tolerable dose and we felt that the dose
22
used in this study for the Phase 3 study was
119
1
comfortably above the threshold that we needed to
2
achieve down-regulation of Bcl-2, which is a dose
3
just above what this particular patient got. Did
4
that happen in 300 patient? We
don't have that
5
information. The willingness of
patients to be
6
serially sectioned for us to obtain this
7
information on a fresh basis is rather limited and
8 it
was simply not part of the study. It
9
overwhelmed our capabilities in year 2000 and was
10 not
done.
11
DR. PRZEPIORKA: If Dr. Tolcher is
here, I
12
have a question. In the in vitro
studies is there
13 a
threshold amount of Bcl-2 that needs to be
14
down-regulated to in order for the chemotherapy to
15
show synergy?
16
DR. TOLCHER: That is a very good
question
17 and
it is not well addressed. Most of the
models
18
are, you know, somewhat artificial and in vitro
19
versus in vivo really has no strict correlation.
20 We
functioned for a period of time with the
21
assumption that 1 mcg/mL is probably the minimum
22
effective concentration. In
almost all of the
120
1
studies published to date we have a steady state
2
concentration of 5 mcg/mL as an average.
So, based
3 on
the work that was done preclinically, published
4 by
Martin Gleave and others, we are well above what
5 we
would need in the in vitro setting but, again,
6 the
major caution always is that it is hard to
7
relate what are the necessary concentrations in
8
vitro to what are the necessary plasma
9
concentrations for maximal effect.
Does that
10
answer your question?
11
DR. PRZEPIORKA: I guess I was
asking what
12 is
the amount of Bcl-2 intracellularly that we need
13 to
get the level down to in order to see the
14
synergy with chemotherapy.
15
DR. TOLCHER: An excellent
question. You
16
know, the issue is that it is dynamic so one
17
doesn't know necessarily. You are
lowering it so
18
that you essentially are shifting the equilibrium
19 in
favor of apoptosis. You clearly do not
need to
20
extinguish all the Bcl-2 to have a pronounced
21
effect in vivo. In fact, you
probably only have to
22
drop it below some threshold and that threshold is
121
1
unknown. It gets more complex as
well in that
2
there is a diversity of Bcl-2 expression in
3
different tumors.
4
So, what I would say is that
it is not
5
necessarily a simple equation where you have to
6
drop it below X amount. It may be
very dependent
7 on
the chemotherapy that is given with it.
So, it
8 is
not clear. The certainty is that we do
know
9
that you do not have to extinguish all the Bcl-2 to
10
have a synergistic effect preclinically.
11
DR. PRZEPIORKA: Thank you. Dr. Bishop?
12
DR. BISHOP: I am relatively new
to all
13
this so I don't know if this question is
14
appropriate or not but I am going to turn it to Dr.
15
Kirkwood and Dr. Haluska. You
made passionate
16
pleas for the treatment of metastatic melanoma in
17
this randomized study. So, would
this treatment,
18
Genasense plus DTIC, become the standard of care in
19 the
control arm for future CALGB and ECOG studies
20
respectively?
21
DR. HALUSKA: I think that is a
reasonable
22
proposition. I think that the
context of this
122
1
trial's conduct is that we have never shown any of
2
these improvements and I think we shouldn't lose
3
site of the fact that we are chipping away, as has
4
been articulated, at numbers that have not been
5
able to be chipped at away before because they
6
haven't existed. So, I think that
that is a
7
decision to be made by the community, but an
8
improvement clinically like we have seen should be
9 the
standard against which other stage 4 therapies
10
will be compared. I think that is
reasonable.
11
DR. BISHOP: Let me make it more
specific
12
then. In your future randomized
trials will this
13
become the control arm? The data
with DTIC we know
14 is
not very impressive yet that is the community
15
standard outside of immunotherapy.
So, as you plan
16
your future trials, and you believe these results
17 are
impressive enough, will that become the control
18
with which new therapies will be developed and
19
compared to?
20
DR. HALUSKA: I wish we had new
therapies
21 to
compare to now. I would have to say that
it is
22
hard to view the future when those new therapies
123
1
become available. The landscape
for drug
2
development for melanoma right now includes other
3
targeted therapies. None of them
is at the stage
4
where we would choose a comparison arm like this
5 but
the short answer to your question is yes.
6
DR. PRZEPIORKA: Dr. Kirkwood?
7
DR. KIRKWOOD: I agree with
Frank's
8
conclusion so I think this is an incremental
9
advance. I think this is
something that we have
10
been trying to do in the studies that I reviewed
11 and
have not succeeded to do. Obviously, if
one
12
were going to take survival as an endpoint in a
13
future study it could still be dacarbazine but I
14
think that we are talking here about response rate
15 and
we don't have anything that has reliably before
16
shown response rates and complete response rates
17
incrementally advanced as this has, with the single
18
exception of high dose IL-2, which we have spoken
19
about previously.
20
DR. HALUSKA: Something else
occurs to me.
21 I
don't think it is the agency's job to support our
22
research endeavors strictly. I
mean, their job is,
124
1 as
I understand it, to make agents available for
2
public consumption. But, clearly,
these decisions
3 do
affect our research and we have, for reasons
4
that are not clear to any of us who work in
5
melanoma, been very unsuccessful in improving
6
overall survival. I don't believe
that as long as
7 we
hold that out as the only endpoint that we can
8
meet that we are going to meet it because it has
9
been such an impediment. But
there is nothing in
10 my
mind that prevents small improvements in these
11
sorts of endpoints from accumulating with addition
12 of
different agents and you can envision a variety
13 of
other things that you could add Genasense to
14
that might also prove additive to the responses and
15
progression-free survival we have seen today.
16
Ultimately, that is how I think we are going to
17
make real progress with the survival endpoint in
18
this field.
19 DR. PRZEPIORKA: Dr. Redman?
20
DR. REDMAN: Thank you but Dr.
Kirkwood
21
answered my question.
22
DR. PRZEPIORKA: Other questions
from the
125
1 committee?
Dr. Tolcher, could you please come back
2 to
the microphone? We need to have you
identify
3
your affiliation, please, for the record.
4
DR. TOLCHER: Sure. I came actually today
5
without personal compensation by Genta or any of
6 the
pharmaceutical sponsors, although my travel
7
arrangements have been paid for Genta.
I have been
8 the
principal investigator on three clinical
9
studies and have acted as an occasional advisor to
10 Genta
and Aventis and have been compensated with
11
honoraria for those less than $10,000.
12
DR. PRZEPIORKA: Thank you. Hearing no
13
other questions, we will break for ten minutes and
14
return at 10:40 to begin the open public hearing.
15 We
will need to begin the afternoon session on time
16 so
please be on time for the next part.
17
[Brief recess]
18 Open Public Hearing
19
DR. PRZEPIORKA: If we could have
the
20 doors
closed, please, we will begin the second half
21 of
this session. This is the open public
hearing
22 and
we actually had many individuals who wanted to
126
1
speak this morning and, in order to give everyone
2 who
is registered a chance to participate and to be
3
fair to all, we will be following some fairly
4
strict procedures. We have a
timer. Each speaker
5 has
been allotted two minutes and at the end of the
6 two
minutes we will ask that speaker to return to
7
their seat and the next speaker to immediately
8
begin. Due to considerations of
fairness and these
9
restrictions of time, only speakers who have
10
registered will be allowed to come to the podium.
11
Both the FDA and the public believe in a
12
transparent process for information gathering and
13
decision-making. To ensure such
transparency at
14 the
open public hearing session of the advisory
15
committee meeting, the FDA believes that it is
16
important to understand the context of an
17
individual's presentation. For
this reason, the
18 FDA
encourages the open public hearing speaker, at
19 the
beginning of your written or oral statement, to
20
advise the committee of any financial relationship
21
that you may have with the sponsor, its product
22
and, if known, its direct competitors.
For
127
1
example, this financial information may include the
2
sponsor's payment for your travel, lodging or other
3
expenses in connection with your attendance at the
4
meeting. Likewise, the FDA
encourages you, at the
5
beginning of your statement, to advise the
6
committee if you do not have any financial
7
relationships at all. If you
choose not to address
8 the
issue of financial relationships at the
9
beginning of your statement it will not preclude
10 you
from speaking.
11
Thank you all for your participation in
12
this portion of the meeting, and our first speaker
13 is
Gail Graham, who is chairman and president of
14 the
William S. Graham Foundation for Melanoma
15
Research.
16
MS. GRAHAM: Good morning. Yes, I am
17
chair and president of the
William S. Graham
18
Foundation for Melanoma Research.
We are widely
19
known as the "Billy" Foundation. Please also note
20
that I am here to represent not any particular
21
therapy or pharmaceutical company though in the
22
past we have accepted financial donations to our
128
1
programs at the Foundation from Chiron, Maxim,
2
Genta, Antigenics and Schering.
However, I have
3
paid my own expenses in order to address you here
4
today.
5
The phone rang and I answered a call that
6
would change my life and the life of our beloved
7
family. Over ten years ago a
doctor called our
8
home and told us that our beloved son had stage 4
9
melanoma. "Mrs. Graham, your
son has three to six
10
months to live." That was
the beginning of my
11
journey into every mother's nightmare, watching
12
your only son disappear before your very eyes.
13
I was told then, ten years ago, that there
14
wasn't anything that could be done for him and no
15 one
prepares you on how to tell your child that
16
there is no hope, nothing that could even extend
17 his
life for an extra month or two.
18
Now, ten years later, what has truly
19
happened to give patients new hope?
What do you
20 say
to patients and their families now? We
want
21
patients to have choices, choices from the onset of
22
their diagnosis not as a second matter of recourse.
129
1
Over those ten years, over 300,000 people have been
2
diagnosed with malignant melanoma in the United
3
States and have had to face that diagnosis and have
4
extremely limited offerings available to them for
5
treatments, and it is long past time that something
6 be
done to offer hope, the hope that they deserve.
7
I am here also to represent the dozens of
8
phone calls that we get on a daily and monthly
9
basis...
10
[Audio system malfunction]
11
DR. PRZEPIORKA: I am sorry, but
thank you
12
very much for your comments. R.M.
Sutton please.
13
MR. SUTTON: No financial
involvement. I
14 am
of clinical relevance--I am free, I am alive, I
15 am
here after my doctor gave me about a month and a
16
half and because of prior medical problems no
17
treatment available, but this trial which has
18
blessed me with time to spend with my son, my
19
daughter-in-law, my daughter, my son-in-law. With
20 all
due respect, should my doctor have waited a
21
thousand or so years until all the kinks were
22
worked out? If we were licensing
aviation today,
130
1
would we have to wait for the law of gravity to be
2
revealed to be assured that we would never fall
3
from the sky?
4
I am 77. I expect to live another
23
5
years. My mother died at 99. I want to see, among
6
many other things, my granddaughter get married and
7
eventually greet my great grandchildren.
I pray on
8
bended knee you approve it so others like me who
9
have been diagnosed with melanoma--thank you, I
10
have a secure place in heaven to join my late wife
11
but, thanks to Genasense, thankfully not just now.
12 You
can give life, hope and achievement. I
hope to
13
write a book on dreams of reality, limited only by
14 my
imagination, inspiration and time. Thank
you.
15
DR. PRZEPIORKA: Thank you, Mr.
Sutton,
16
very much. Davie Bernstein, please.
17
MR. BERNSTEIN: My name is David
18
Bernstein. I paid my own way
here. I have taken
19
time off from work in order to address you here
20
today. I am 51 years old, a
husband, father of two
21
little girls, a fourth grade teacher in New Jersey.
22 Two
years ago I was diagnosed with stage 4 melanoma
131
1
after discovering a lump in my chest.
We were
2
devastated. We had found the
disease had already
3
spread to my lungs.
4
I sought a group at Thomas Jefferson
5
University Hospital in Philadelphia to be treated.
6 We
discussed various options for treatment, all of
7
which included going on various forms of
8
chemotherapy. I learned that DTIC
was the standard
9
care although it was described as having very
10
limited results. My doctor also
told me about a
11
clinical trial they were conducted for a drug
12
called Genasense. I qualified for
the trial,
13
feeling oddly lucky that my tumor was large enough,
14 and
received Genasense with DTIC.
15
Genasense was administered through an
16
automatic pump that I wore like a fanny-pack for
17
five days, followed by a one-hour infusion of DTIC.
18
After six weeks, or two treatment cycles, I got a
19 CT
scan to monitor the size of my tumor.
The scan
20
showed that my tumor had already begun to shrink.
21 I
remained on the therapy for a total of 16
22
treatments and was scanned every six weeks, each
132
1 one
coming back clear of tumors. Throughout
my
2
treatment, I was very well supported by the team at
3
Thomas Jefferson that included my oncologist, Dr.
4
Sato, and Tracy Newhalls, the clinical liaison.
5
I stopped treatment in August, 2003 and
6
have remained tumor-free since then.
I am here
7
today because I received Genasense in this study.
8
Genasense now needs to be made available to the
9
thousands of people like me who have received or
10
will have received the diagnosis of advanced
11
melanoma. People need to know
that there is hope
12 for
this disease in the form of new drugs.
13
Genasense worked for me and others should have the
14
same chance I did. Thank you.
15
DR. PRZEPIORKA: Thank you very
much for
16
your words. Erica Weiss, please.
17
MS. WEISS: Good morning. My name is
18
Erica Weiss and I am the director of patient
19
education and outreach for the Wellness Community.
20 For
the record, The Wellness Community will receive
21 an
unrestricted educational grant from Genta and
22
Aventis. However, I received no
compensation for
133
1 my
presence here today.
2
By way of background, the Wellness
3
Community is a national non-profit organization
4 that provides free services for people with
cancer
5 by
way of support, education and hope. Our
6
programs include professionally facilitated support
7
groups, educational programs on nutrition, mind,
8
body--programs like this. We aim
to help people
9
affected by cancer regain a sense of control over
10
their lives, feel less isolated and restore a sense
11 of
hope for the future regardless of the stage or
12
type of their disease. Last year
we served about
13 30,000
people with cancer, including people with
14
melanoma.
15
At the Wellness Community we have learned
16 a
great deal from the people we serve and we really
17
value the importance of an educated and empowered
18
patient, and since we feel that people with cancer
19
often feel stigmatized, alone and overwhelmed with
20
grief, they feel stronger and more hopeful when
21
they have more options available for their disease.
22
When a cancer like melanoma results in 80
134
1
percent of skin cancer deaths and when limited
2
treatment is available for advanced melanoma, it is
3
clear that we are in great need of new treatment
4
options and better access to those treatments. At
5
this time we have the opportunity to expand the
6
chance that these families have in their daily
7
fight for life and we feel strongly about
8
supporting that opportunity, assuming that the
9
treatment promise has manageable side effects,
10
assuming there is progression-free survival time,
11
even if only for a few weeks or months, and other
12
positive outcomes.
13
I ask today that you carefully consider
14 the
plight of people with melanoma and understand
15 the
range of both physiological and psychological
16
issues that they face daily.
Please take a
17
leadership role in considering the approval for a
18
broader range of treatments based on sound science
19 and
answers to hard questions, and then encourage
20
patients to be informed, empowered and possibly
21
optimistic about the potential for a longer,
22
healthier life. Thank you.
135
1
DR. PRZEPIORKA: Thank you very
much. Dr.
2
Anna Pavlick, please.
3
DR. PAVLICK: Good morning. Thank you for
4
allowing me to address the committee.
I am one of
5 the clinical investigators on this trial. I
have
6
received no financial compensation for coming down
7
here, however, I do receive research support
8
through Genta and Aventis.
9
I am actually here on behalf of my
10 patient.
This is Mrs. Kovati. Mrs. Kovati
was my
11
first patient to be enrolled on the Genta trial in
12 my
institution. She was told by a few other
13
melanoma oncologists that she had six months to
14
live and there were no options for her.
She came
15 to
me four and a half years ago in a wheelchair,
16
with a leg full of melanoma, large pelvic
17
adenopathy and multiple tumors in her abdomen and
18
said, "I'm only 56 years old.
I don't want to die.
19
Help me." I explained to her
that we had this
20
clinical trial available to her and told her
21
full-well I was not sure if this was going to help
22
her, however, we knew what her alternative was, so
136
1 she
went on study.
2
She was featured in CURE magazine last
3
summer because, I am proud to say, Mrs. Kovati had
4 a
complete response. She now remains three
and a
5
half years out of therapy in a continued complete
6
response; has been able to get out of her
7
wheelchair. She no longer walks
with any assistive
8
devices. She was able to dance at
her son's
9
wedding a year and a half ago, and she was unable
10 to
come down here today to be with us because she
11 is
now experiencing the birth of her grandchild,
12 the
first one that she thought she would never-ever
13
see.
14
I felt it was on her part and on the part
15 of
all the other melanoma patients that I treat
16
that I needed to come down here and tell you what a
17
wonderful experience it has been for me to work
18
with this new drug that truly holds hope for
19
patients who have absolutely no options.
Thank
20
you.
21
DR. PRZEPIORKA: Thank you very
much. Dr.
22
Lawrence Green, please.
137
1
DR. GREEN: I have no financial
2
disclosures to report.
3
[Slide]
4
My name is Lawrence Green. I am a
5
dermatologist and dermosurgeon in private practice
6 in
Montgomery County. I also teach a weekly
7
dermosurgery clinic at George Washington University
8 to
the dermatologist residents.
9
I am here today as a professional member
10 of
the Skin Cancer Foundation specifically because
11 I
have an interest in skin cancer.
12
[Slide]
13
Skin Cancer Foundation is the only
14
national organization that is non-profit, dedicated
15
solely to eradicating the world's most common
16
malignancy, which is skin cancer and it has been
17
around for 25 years, educating the public, among
18
other things. Despite these
ongoing efforts, as
19 you
know, the incidence of skin cancer, especially
20
melanoma, continues to rise at an alarming rate.
21
[Slide]
22
One in three cancers this year will be
138
1
skin cancer which translates to 1.3 million new
2
cases of skin cancer in the United States this
3
year. Basically, that means that
20 percent of the
4
population in the United States will develop skin
5 cancer
in their lifetime.
6
[Slide]
7
One person dies every hour from melanoma.
8 In
fact, if you look at it, melanoma is basically
9 the
most common cancer in women between the ages of
10 25
and 35.
11
[Slide]
12
In light of these abysmal statistics, it
13 is
painfully clear that providing public education
14
messages on sun protection, skin cancer prevention
15 and
early skin cancer detection is not enough.
The
16
Skin Cancer Foundation is speaking here today, and
17 I
am speaking on behalf of it, as part of its
18
patient advocacy mission to support skin cancer
19
research and the latest advancements in effective
20
treatments for its constituents.
21 [Slide]
22
Sadly, there are currently very few
139
1
effective treatments available for late stage
2
melanoma patients. Therefore, if
this new
3 treatment
shows promise, on behalf of myself and
4 The
Skin Cancer Foundations, the many patients and
5
their families who have been affected by melanoma,
6 we
encourage this committee to carefully consider
7
it. Thank you.
8
DR. PRZEPIORKA: Thank you, Dr. Green.
9
Diane Murphy, please.
10
MS. MURPHY: Thank you for
allowing me to
11
come before this scientific panel to urge fast
12
approval for Genta's drug Genasense.
Three years
13 ago
I was diagnosed with stage 4 melanoma, and Dr.
14
Hersh, at the Arizona Cancer Clinic in Tucson, told
15 me
that without treatment statistics would show
16
that I had around nine months to live.
This was
17
shocking news for me because as a family we have
18
been living on organic food, drinking bottled
19
water, exercising, staying away from chemicals and
20
doing whatever else we thought would give us a
21
healthy life. So, how could this
lead to a golf
22
ball sized tumor?
140
1
I was biopsied, diagnosed and, thankfully,
2
referred to Dr. Hersh. I was
considering no
3
treatment at all but Dr. Hersh persevered,
4
suggesting that I was a good candidate for the
5
experimental Phase 3 drug, which I did agree to try
6 if,
for no other reason, although it might not help
7 me
it would help someone down the road.
8
It did help. As my doctor told
me, I have
9 a
complete response to my treatment and can now
10
enjoy celebrating my big 70th birthday, which I did
11 by,
among other things, buying shares of Genta.
12
[Laughter]
13
Hopefully, none of you today making a
14
decision on this drug has ever had friends or loved
15
ones sitting in a chemo treatment room.
It is the
16
saddest and most depressing place to spend time.
17 You
can smell the fear, the misery, hopelessness
18 and
anger, and see the fatigue in all their faces
19
under all the green hats hiding their bald heads.
20
Help for each and every one of the patients is
21
hearing the word "remission" and that is what
22
Genta's drug gave me, and I am here to encourage
141
1 you
to pass this drug for approval.
2
In closing, I want to thank God and the
3
people in my life, my husband Jim who is always
4
there 24/7, for hundreds of prayers from friends
5 and
acquaintances, both known and unknown, Dr.
6
Hersh who truly is a healer in the greatest sense
7 of
the word and my oncology nurse, Cindy who
8
encouraged me to get through each treatment day. I
9
pray that all the poor souls going through this
10
dreadful disease can have the same care, support
11
team and access to the latest drugs such as Genta's
12
Genasense. Thank you.
13
DR. PRZEPIORKA: Thank you very
much, Ms.
14
Murphy. Dr. Asher Chanan-Khan,
please.
15
DR. CHANAN-KHAN: Hi. I have received
16
honoraria for a speaking engagement.
I have
17
received clinical trial support from Genta and have
18 not
been compensated for anything for today's
19
meeting.
20
I would like to thank the committee for
21
allowing me to voice my opinion in the matter of
22
Genasense. I come here from
Russell Park in
142
1 Buffalo,
New York, where I am entrusted with the
2
care of patients with multiple melanoma and chronic
3
lymphocytic leukemia. I am one of
the clinical
4
investigators involved in the studies exploring the
5
role of Genasense in these incurable and rather
6
frustrating diseases.
7
The NCI identified these as orphan
8
diseases, thus, emphasizing the need for developing
9 new
and novel therapeutic options. Based on
my
10
personal experience as a clinician and as an
11
investigator, I am able to comfortably state that
12 the
agent is safe and well tolerated during these
13
clinical trials that I am conducting.
No long-term
14
side effects in the patients that I have treated
15
have been noted. In fact, with
this drug a number
16 of
patients with CLL and multiple melanoma have
17
benefited clinically and continue to benefit as of
18
today.
19
In conclusion, I therefore feel that this
20 is
a safe drug with a predictable and manageable
21
side effect profile, and it does bring hope to a
22 lot
of patients in my clinic who are facing an
143
1
incurable cancer. Thank you.
2
DR. PRZEPIORKA: Thank you very much. Dr.
3
Tolcher, please.
4
DR. TOLCHER: I am a medical
oncologist in
5 a
cancer therapy research center. I have
given my
6
disclosures already. I am an
investigator with one
7 of the larger clinical experiences with
oblimersen,
8
having treated 63 patients in 288 courses of
9
oblimersen during the conduct of 3 clinical
10
studies. This includes one
patient who received
11 the
maximum of 25 courses of this agent.
12
The toxicity profile of oblimersen is
13
modest and largely predictable.
The majority of
14
adverse events experienced by patients are related
15 to
the chemotherapy itself and, again, are
16
predictable for that chemotherapy agent.
They do
17 not
require any special management above that of
18
what a standard medical oncologist provides.
19
For those toxicities that can be
20
attributed to oblimersen alone, they include a
21
transient lymphopenia, pyrexia that occurs during
22 the
infusion but can be treated with standard
144
1
antipruritics, and complications of the central
2
venous catheter. Patients with these
toxicities
3 can
be safely retreated with the agent without
4
evidence of cumulative increases or increases in
5 the
severity of these toxicities.
6
Interestingly, and I think really
7
importantly, patient acceptance of the oblimersen
8
treatment and its inherent cumbersome pump is high
9 due
to the low incidence of adverse events
10
associated with oblimersen. From
a clinical
11
perspective, oblimersen can be safely and feasibly
12
administered to patients with cytotoxic
13
chemotherapy over many multiple courses.
Thank
14
you.
15
DR. PRZEPIORKA: Thank you, Dr.
Tolcher.
16 Dr.
Patrick Cobb, please.
17
DR. COBB: Patrick Cobb, I am
medical
18
oncologist from Montana. I
receive research grants
19
from both Aventis and Genta. I
have not been
20
compensated for my time.
21
We have participated in a trial of
22
Genasense in CLL and I will address some of the
145
1
safety concerns about it. We have
treated three
2
patients with this. All these
patients had disease
3
refractory to fludarabine chemotherapy.
One
4
patient received six courses of this and had no
5
toxicity greater than grade 2 and remains in
6
complete remission two years later.
Another
7
patient was treated with the same regimen and had
8 an
Aspergillus lung infection at the beginning of
9 his
course and went into complete remission after
10
only one course and continued in complete remission
11
after two years. He relapsed a
while back and is
12
receiving another course of Genasense now.
13
In summary, we found Genasense to be a
14
very well tolerated drug when it was given to our
15
patients with chronic lymphocytic leukemia. As a
16
clinical oncologist I see a lot of patients with
17
metastatic melanoma and, as you have already heard
18
this morning, there are very limited options for
19
their treatment and we need more treatment options.
20
From the data we have seen presented today, it
21
appears that Genasense is both a safe and an
22
effective drug. Thank you.
146
1
DR. PRZEPIORKA: Thank you, Dr.
Cobb.
2
Harrison Blanton, please.
3
MS. BLANTON: Betty Blanton, from
Shelby,
4
North Carolina. I came at the
request of my
5
oncologist, with no compensation but I have
6
discussed travel expenses with Genta.
7
I came to Carolina Regional Medical Center
8 in
Charlotte in October, 1995 after my melanoma
9
reappeared following two previous melanoma
10 surgeries. Later in my treatments as the disease
11
progressed surgery was no longer a viable option.
12
When your oncologist tells you that you have
13
metastasized melanoma for which there is no
14
surgery, thankfully, my family and I considered the
15
best course and we decided that to be the Genasense
16
trial, as was suggested by Dr. Gary Fernad of
17
Carolina Health System.
18
I began with the trial in January, 2003
19
with eight cycles. My last was in
July, 2003. My
20
gratitude goes to my three sons who have provided,
21 and
still do, transportation since I live an hour
22
from Charlotte. I received the
Genasense
147
1 continuously
for five days and then would go back
2 for
my DTIC. The Genasense treatment was not
a bad
3
experience, although a little trying to dress and
4
keeping the wires intact was something interesting
5
which I am sure the women can relate to.
During
6
that time I was referred to by my friends as the
7
lady with the fanny-pack.
8
On days five of the Genasense treatment I
9 did
go back to Charlotte and received my DTIC.
If
10 I
followed the medication for nausea as directed, I
11 was
able to function normally all the time.
There
12
were times when anemia was a problem but this was
13
addressed by the doctor and his team.
Sometimes a
14
transfusion was needed but on most days I was able
15 to
do my normal office work in the mornings as a
16
church secretary and teach piano in the afternoons,
17
both of which I have enjoyed for over 50 years now.
18 On
Sundays I play the organ at the church.
Out of
19
those eight cycles of treatments only one Sunday I
20 was
not able to play.
21
I have nine grandchildren and two great
22
grandchildren. They are, indeed,
my life as each
148
1 of
you share with your families. But I am
here
2
today because, I believe, the Genasense trial was a
3
success for me. I am still able
to work, enjoy my
4
family and continue to live independently, and it
5 is
my hope that this experience will have an impact
6 on
the lives of others who know melanoma
7
personally.
8
DR. PRZEPIORKA: Thank you very
much. Dr.
9
Jonathan Lewis, please.
10
DR. LEWIS: Distinguished members
of the
11
committee, good morning. My name
is Jonathan
12
Lewis. I come before you wearing
two hats. For
13
more than eight years I worked as a surgical
14
oncologist at Sloan-Kettering.
Although I still
15
follow patients, the second hat I wear is
16
developing cancer drugs in the context of a private
17
start-up company. I have no
financial interest at
18 all
in Genta. They have not paid me
anything; they
19
have not asked me to be here.
Their CEO, Ray
20
Morrell, referred many melanoma patients to me
21
while we both worked at Memorial.
I have only had
22
sporadic contact with him for several years; I have
149
1 not
spoken to him for at least six months.
2
I speak to you today because this
3
committee's decision is important in the context of
4
both the science and art of treating melanoma
5
patients and the science and art of cancer drug
6
development. I have been involved
in the care of
7
thousands of melanoma patients at Memorial. I have
8
treated well over a thousand, and I have also
9
conducted and been part of many experimental
10
clinical studies in this disease.
11
As we have heard, stage 4 melanoma is an
12
extraordinarily difficult problem.
As I interpret
13
these data presented today, it strikes me that
14
despite the fact that the study clearly missed the
15
statistical primary endpoint, every single
16
analysis, including response rate, progression-free
17
survival and survival demonstrates an advantage for
18
those patients receiving the test agent.
I
19
understand that statistical improvement in survival
20 is
the gold standard but I am, nonetheless, very
21
focused on the observation that Genasense shows
22
effectiveness in the setting of a hundred percent
150
1
lethal disease. In the context of
the disease, all
2 of
these are very likely to be clinically
3
meaningful.
4
I am here today in part because a patient
5 of
mine with stage 4 melanoma is sitting in the
6
audience. He is a highly
decorated, allegedly
7
retired senior FBI agent who has served this
8
country extraordinarily. His care
has involved a
9 lot
of the science and art. I have been
through
10 the
data with him and he has a tremendous amount of
11
common sense, wisdom and understanding and, on
12
reviewing these data, he asked me how can this drug
13 not
be approved. I am grateful for your
time.
14
Thank you very much.
15
DR. PRZEPIORKA: Thank you. Cathy
16
Liebermann, please.
17
MS. LIEBERMANN: Good
morning. My name is
18
Cathy Liebermann and I am a two-time cancer
19
survivor. I am here with my
daughter Lisa and her
20
husband Aaron to share our family struggle with
21
melanoma.
22
After reading about this meeting last week
151
1 in
the Wall Street Journal we felt obligated to be
2
here today, and we paid our expenses to do so. Our
3
story begins in 1996 when I was undergoing
4
chemotherapy for Hodgkin's disease.
My husband
5
Mark's primary concern at that time was my
6
treatment and helping me with my battle.
All the
7
while Mark ignored a growth on his scalp. Because
8 the
growth was pink and perfectly round, Mark did
9 not
think it urgent to see a doctor.
However,
10
months later he was diagnosed with melanoma and the
11
lesion was removed. We were
elated that the
12
pathology results showed no disease in Mark's lymph
13 nodes
and no further treatment was needed.
14
Six years later, in February 2003,
15
metastatic melanoma was confirmed.
We sought the
16
advice of experts that included Dr. John Kirkwood
17 and
a family friend, Dr. Jerome Groupman, who
18
referred us to Drs. Michael Atkins and John
19
Richards. Mark then proceeded
with four cycles of
20
biochemotherapy. In July he
walked down the aisle
21
with Lisa at her wedding.
22
Only two months later tumors began to grow
152
1
again. It was then that Genasense
was recommended
2 to
us. We were disappointed when Dr.
Richards
3
informed that the Genasense trial was no longer
4
enrolling patients so Mark began other treatment
5
instead in November. On January
10th Mark died at
6 the
age of 54.
7
There is no way to know if Genasense would
8
have helped Mark but based on the trial results I
9
believe that had Mark taken this drug he might be
10
standing here with us today.
Lisa, Aaron and I are
11
here to plead with you to vote in favor of
12
Genasense for all those who suffer with this
13
disease and for their families who just want a few
14
more days, weeks or months with their loved ones.
15
Thank you for listening.
16 Committee Discussion
17
DR. PRZEPIORKA: I have no other
18
individuals registered. I do want
to apologize on
19 behalf of the committee to Ms. Graham for the
sound
20
going off before she completed her statement. We
21
have asked if she wished to make any additional
22
comments and I understand she does not.
If you
153
1
need to change your mind now, please feel free.
2
Otherwise, we will go on with the rest of our
3
meeting but we do apologize to Ms. Graham.
4
The next item on the agenda is the
5
questions posed from the FDA to the committee. We
6
have all received these previously.
They include a
7
rather lengthy prologue which Dr. Pazdur has chosen
8 not
to review for us. So, we can go straight
to
9
page three and we will be voting on questions one,
10 two
and three and question four is for discussion
11
only. Let me start with
question number one,
12
given the thrombocytopenia concerns noted above,
13
does the committee believe that the small observed
14 differences
in the response rates, that is, less
15
than 5 percent, and in progression-free survival,
16 the
difference in median days between arms of 13
17
days with a p value of 0.006, represent real
18
effects of Genasense when added to DTIC?
19
I am going to ask for discussion for a few
20
minutes before we actually go around and take a
21
vote. So, if anybody has any
comments on this
22
question, please feel free.
154
1
DR. GRILLO-LOPEZ: I have a point
of
2
order. I think the question needs
to be worded
3
differently because the way it is worded it is
4
biased towards the analysis of the data by the FDA.
5 I think
we need to consider as a committee both the
6
FDA's analysis as well as the sponsor's analysis.
7 So,
I would say that the qualification of the
8
differences as small should be taken out and the 13
9
days, which comes from the FDA analysis, should be
10
taken out.
11
DR. PRZEPIORKA: Dr. Pazdur, do
you accept
12 the
changes in your question, or Dr. Temple?
13
DR. TEMPLE: The committee
obviously is
14
supposed to consider all the data it heard. It
15
heard more than one assessment of both of those
16
things and, obviously, can consider both.
17
DR. PAZDUR: I share that, and as
I
18
pointed out in my initial comments, I think what
19 one
has to take a look at is the individual
20
contribution that the drug is making.
Remember, we
21 are
dealing with a combination of a drug so one has
22 to
take a look at the delta also.
155
1
DR. PRZEPIORKA: Thank you for
2
accommodating this need so a more unbiased question
3
perhaps would be, given the concerns noted above,
4
does the committee believe that the observed
5
differences in response rate and progression-free
6
survival represent real effects of Genasense when
7
added to DTIC? Dr. D'Agostino?
8
DR. D'AGOSTINO: I noted that the
second
9
question picks up the ordering of the analysis.
10 Are
we supposed to take question one as if we use
11
some sort of clinical judgment, are these effects
12
substantial, ignoring the fact that we may not be
13
able to attach any statistical significance to
14
them?
15
DR. PRZEPIORKA: The answer would
be yes.
16 Dr.
Hwu?
17
DR. HWU: I would like to review a
little
18 bit
the background of the treatment of advanced
19
metastatic melanoma. In the last
30 years we have
20
made very small progress. The single-agent
21
chemotherapy gradually evolved into the combination
22
chemotherapy and also the development of
156
1
immunotherapy and the combination of a
2
biochemotherapy involving the interferon
3 interleukin and the chemotherapy. That evolvement
4 is
primarily based on the findings of the pilot and
5
Phase 2 studies. Those trials
have clearly
6
demonstrated that when you combine several agents
7 the
response rate definitely increased, in some
8
cases double or triple, especially with
9
biochemotherapy. Yes, the price
you pay is very
10
high; it is toxic. However, in
the Phase 3 trials
11
none of those combination therapies has
12
demonstrated that even with the response rate the
13
difference is clinically significant but there is
14 no
impact on the outcome of the survival, not
15
statistically significant.
16
So, in year 2002 we started the AJCC
17
staging system which clearly separates the patients
18
with stage 4 disease into three prognostic groups,
19
M1a, which has disease in the skin and the lymph
20
nodes; M1b, which can have soft tissue and the
21
lymph nodes but also has lung metastasis; and M1c
22 is
the patients who have visceral disease other
157
1
than lung or with elevated LDH.
The reason those
2
patients were categorized in three groups is really
3
based on their survival. The data
is from over
4
1000 patients from nine major cancer centers.
5
Irrespective of what their treatment was, the
6
median survival for M1a is 16 months; the M1b group
7 is
14 months because their survival is correlated
8
with M1a for the first year and then that becomes
9
consistent with the M1c group.
The M1c group has
10 the
shortest median survival of 7 months or less if
11 you
have brain metastasis which is less than 6
12
months.
13 So, clearly, if we want to make any
impact
14 on
the survival of the patients with stage 4
15
disease we have to make the treatment more
16
effective for the M1c group. I
have to
17
congratulate the sponsors of this study that they
18 did
not exclude the patients with M1c which is a
19
very, very bad group. However, it
was not balanced
20 on
the two arms. The M1c group has more
patients,
21 253
on the DTIC alone group--257, and in the
22
experimental group there were 226.
The imbalance
158
1 was
also seen in stage M1a. On DTIC it was
50
2
patients and the experimental arm had 61 patients.
3
So, what is the outcome when you compare
4
that everybody is getting the DTIC and only the
5
experimental arm is getting the experimental drug?
6 So,
which group benefits the most by adding the
7
experimental drug? It is not
surprising to see
8
that most of the patient benefit is with the M1a
9
group because it was clearly shown in the previous
10
Phase 1/2 trial that patients who had responded
11
well to the Genasense plus DTIC is the group with
12
lymph node and also skin metastases.
So, in this
13
study the M1a group in the experimental arm--13
14
patients had a response, objective response as
15
compared to DTIC with 6 patients.
16
In the M1b group 16 out of 96 patients
17
responded to the experimental group and 9 out of 75
18 in
the DTIC alone group. However, in M1c 16
out of
19 226
patients responded to the experimental drug as
20
compared to 11 out of 227 of DTIC alone.
21
So, I definitely say yes, there is
22
activity of this drug when it is compared with
159
1
DTIC. Are we going to make any
difference in
2
prolonging survival of our patients?
Believe me, I
3
desperately want to have some drug that can help
4
with my patients. After 15 years
in this field I
5 cry
every time when I lose a patient; I feel it is
6 a
personal defect. But, unfortunately,
this drug
7 is
not the answer, at least the way it is
8
administered. We are helping the
best prognostic
9
group of patients and I hope that with continued
10
effort we will eventually help the group of M1c
11
patients. Thank you.
12
DR. PRZEPIORKA: Dr. Cheson?
13
DR. CHESON: Yes, first of all to
14
follow-up on what you were saying, it is clear that
15
with these biotherapeutics, or however we
16
categorize this drug, that we don't have a clue as
17 to
the optimal way to use them. We base it
on cell
18
lines, pharmacodynamic things, but that doesn't
19
mean that this is the best way to do it.
My
20
concern is that if we consider this unapprovable
21 the
drug is going to die and we will never figure
22 out
how to use it, and how to apply it better, and
160
1 how
to study it better in other diseases as well as
2
melanoma, melanoma being one of the two diseases
3
increasing in frequency; the other being
4
lymphoma--we have to get our plug in there.
5
The other point I want to make is that I
6 sat
here a few months ago at another ODAC meeting,
7 and
this was mentioned earlier, and saw another
8
drug approved with a response rate for which the
9
lower limits of the confidence interval was 5.4
10
percent with two huge negative Phase 3 trials
11
without even a twinkle of progression-free
12
survival, without any suggested difference of
13
long-term survivors. To me, these
results are a
14 lot
more encouraging than that drug that was
15
approved at a prior meeting. And,
that is all I
16
have to say about point number one.
17
DR. PRZEPIORKA: Dr. D'Agostino?
18
DR. D'AGOSTINO: Why will the
drug die?
19 You
don't think the company will pick it up with
20 the
promising results here? The studies are
too
21
expensive?
22
DR. CHESON: You know, I have no
161
1
conversations with the company about that or
2
anything else but with a small company that has
3
devoted a lot of resources into a particular drug,
4 if
it doesn't get approved then, based on economics
5
etc., drugs tend to fade away.
6
DR. PRZEPIORKA: Dr. Temple?
7
DR. TEMPLE: Not to state the
obvious, but
8
really we need to know from you whether you think
9 it
works, not whether you feel bad for the company
10 or
feel bad for the state of oncology development.
11
DR. CHESON: No, that is not the
point. I
12 do
think it works. I think there is a
strong
13
signal here but I think, as with that other drug,
14 we
don't know the optimal way to use it.
But there
15 is
a signal here. I do believe the
16
progression-free survival data, as we will get to
17 in
the next point. This committee discussed
last
18
time, and may discuss tomorrow, that
19
progression-free survival may perhaps be the better
20
endpoint and, had this trial been started today
21
instead of several years ago, they would have been
22
recommended to use progression-free survival and we
162
1
might not have been having this sort of discussion.
2
DR. TEMPLE: But this question is
about
3
whether you believe there is a difference in
4
progression-free survival. The
importance of it
5
really is what the second question is.
6
DR. CHESON: Well, I will vote yes
on that
7
when it comes to my time to vote.
8
DR. TEMPLE: Okay. Even though the
9
question has been modified appropriately because we
10
don't want to put bias in it, you do need to tell
11 us
what you think of the various comments that
12
various people have made about the difference in
13
time of assessment and whether those shake you or
14
not. That is what this question
is.
15
DR. CHESON: I will leave that to Dr.
16
George who is about to ask a question.
17
DR. GEORGE: I have a number of
comments
18
about this. To me, some of this
is rather
19
disturbing and I guess that is why we have it
20 before
the committee. If it were easy we
wouldn't
21 see
it.
22
The general strategy of when the primary
163
1
endpoint is not met and looking at secondary
2
endpoints is bothersome from a regulatory point
3
view point and scientific view point just on the
4
surface. That is, one way it
could have been
5
done--of course, we wouldn't be talking about this,
6 at
least in the same way if the primary endpoint
7 had
been progression-free survival and more tightly
8
done with the measurements. But,
you know, one way
9 it
could have been done would have been a bigger
10
study, of course, but you could have said, all
11
right, we are going to look at the primary endpoint
12 and
the secondary endpoints and we are going to
13
make adjustments. The adjustments
basically are we
14
have to be more sure of the results, therefore, we
15
have to have a much bigger study. Of course, this
16 is
already a large study.
17
So, getting back to the point, there
18
wasn't an advantage in survival.
There may have
19
been some signal there. That is,
some very small
20
percentage of patients, those who achieve a CR, may
21 be
the long-term survivors and may, in fact, be
22
different in the really long term.
That is, you
164
1
might have--what?--if you look at the survival
2
curves at about 20 months they are identical but
3
there is some evidence obviously both from the
4
testimonials and from the data that there are some
5
patients who are making it beyond that.
6
But to pick up that kind of difference, of
7
course, is very, very difficult and takes huge
8
sample sizes and that is sort of out of the
9
question here. But what is
bothering some people
10
here is that they are thinking there might be
11
something here but it just isn't clear.
12
Just to make my own point on this, it is
13
clear that the overall survival, from a regulatory
14
view point, wasn't significant. I
am very
15
suspicious of the progression-free survival. I
16
didn't get the data myself, of course, and go over
17 all
this but I am very worried by the differential
18
measurement timing and the effect of this, the
19
potential effect of this on attenuating that
20
result, maybe attenuating it down to a point where
21
there is really essentially no difference between
22 the
two.
165
1
So, I am sort of left at looking at these
2
response rates and then I hear that there is this
3
question about whether this independent assessment
4 of
the response rate--there is some question about
5
that and, again, I am not clear on what it all
6
means. It sounded plausible that
maybe if this
7
independent group had had more of the background
8
clinical information it wouldn't have been so
9
discrepant, but the fact is it was discrepant. So,
10 I
am struggling with all these things in the face
11 of
what might be a promising agent but probably at
12 a
very low level.
13
DR. PAZDUR: I just wanted to
emphasize
14 why
we drew up these questions the way we did.
If
15 you
remember my opening comments, we first have to
16
make sure that there is a biological effect. What
17 is
the effect of this drug on the endpoint that we
18 are
entertaining, and then how adequately
19
characterized is that effect? We
have to answer
20
that question first before we go and discuss the
21
clinical relevance because the clinical relevance
22 of
a certain drug brings in the risk-benefit
166
1
relationship and, as I pointed out, benefit cannot
2 be
discussed unless it is adequately characterized,
3 and
this is the sense of the questions and why we
4 are
asking them in the way we are.
5
DR. PRZEPIORKA: I would just then
like to
6 ask
if we could split question 1 into 1A and 1B.
7
DR. PAZDUR: That would be fine.
8
DR. PRZEPIORKA: So, 1A being the
9
difference in response rate is pretty objective and
10 I
think we can address that. I am just
sorry to
11
hear that the study was not designed truly based on
12 the
best way determined in this Phase 1 study, as
13 Dr.
Hwu pointed out earlier, and also that there is
14
really no biological correlate that was looked at,
15
going instead straight from a Phase 1 to a Phase 3.
16 So,
there is a huge number of design issues which I
17
think really limited the difference in response
18
rate that we are seeing here.
19
I have to agree with Dr. George that there
20 is
a tremendous bias ascertainment here with the
21
progression-free survival data and that is why I
22
would like to ask that these two questions be
167
1
answered separately. Dr.
D'Agostino, you had more
2
comments?
3
DR. D'AGOSTINO: In some sense I
was going
4 to
endorse what was said. I mean, we have
to
5
understand, if I am understanding correctly, that
6
these were secondary outcomes we are looking at,
7 and
sort of the way that one would rigorously
8
define these and then ascertain them is somewhat
9
missing. So, I am stuck, as you
point out, with
10 the
difficulty with progression-free survival and
11 how
that can move around depending on some
12
assumptions.
13
I am also concerned with the
response rate
14 in
terms of how rigorous that was. I am
quite
15
surprised that the outside independent group was
16
somehow or other only there for quality control,
17 and
the quality control was somehow or other not
18
able to work because it wasn't given all the data.
19 I
find those aspects of the study to really bother
20 me
in terms of how do we interpret these relatively
21
small numbers.
22
DR. PRZEPIORKA: Dr. Taylor?
168
1
DR. TAYLOR: I guess I have a
concern
2
about progression-free survival in that there are
3
some patients who have very slow growing tumors
4
and, if you are going to use that as a measurement,
5 in
particular people with the soft tissue type
6
disease, I think you have to know how rapidly they
7
were progressing before they were treated, and if
8 you
have someone who had very slow growing disease
9
that might be impacted on that.
10
The second thing that as a clinician I
11
have seen is that melanoma is a particularly
12
unpredictable disease. Although
its response to
13
chemotherapy has been dismal, I have patients whom
14 we
put on tamoxafin studies and who are now 20
15
years out in complete remissions.
So, it makes it
16
very hard for me to not be concerned when I see
17
small numbers of patients getting benefit about
18
whether it is truly the drug or the natural history
19 of
that particular melanoma.
20
DR. PRZEPIORKA: Dr. Bukowski?
21
DR. BUKOWSKI: The issue of
response rates
22 I
think is an important one to consider.
We have
169
1
looked in melanoma, and I believe I am correct
2
here, in randomized trials where we have added
3
biological agents to chemotherapy and have seen
4
increments in response rates in the past that were
5
significantly higher than the chemotherapy alone.
6
Unfortunately, those studies demonstrated no
7
benefit in terms of survival or other secondary
8
effects.
9
So, I think we have to keep this in mind
10 as
we consider this particular drug. There
is an
11
increment in response here that may be a signal but
12 we
have seen this before without the signal of
13
survival being met. Melanoma is
not unique in this
14
situation, obviously, but this is a concern when
15 you
look at response rates and we are saying
16
response is one measure of drug effect here and we
17
have seen this before in this disease.
18
DR. PRZEPIORKA: Before we go on
to the
19
vote, are there any other comments from the
20
committee? Dr. Rodriguez?
21
DR. RODRIGUEZ: I share similar
concerns
22
that have already been voiced with regards to the
170
1 PFS
endpoint and that there clearly was some
2
difference in the timing to assessment of that
3
endpoint.
4
I think as a clinician there is one thing
5
that can't be argued and that is, as I look at this
6
data, the arm that got Genasense clearly had more
7
complete remissions. I am staring
at that and I
8
can't let that go. I mean, we
have seen some of
9 the
survivors here today and one can't argue with
10 the
living.
11
We all know as oncologists that we will
12
never get to a cure unless one gets a complete
13
remission. So, it is intriguing
to me that it
14
seems that this drug probably improves on the
15
quality of response rather than the overall total
16 response or DTIC. The question is what makes the
17
people who did get the complete responses different
18
than the other patients. I am so
disappointed,
19
like Dr. Hwu, that we don't have anything that
20
correlates that will point us to the appropriate
21
patients for whom this drug is indicated.
22
DR. PRZEPIORKA: Dr. Reaman?
171
1
DR. REAMAN: I regret that we have
sort of
2
brought up the past in a prior meeting of this
3
committee but, unfortunately, it has been brought
4 up
and there was a suggestion to approve an agent
5
with a response rate that was of a similar
6
magnitude. I feel that we are
being called upon to
7
make a similar decision again with a hint of a
8
response with an agent that may disappear if it is
9 not
approved at this committee meeting.
10
Also, I am troubled by the fact that the
11
response rates and the methods for independent
12
review were as troublesome in this study, but I
13
just feel like we are between a rock and a hard
14
place in trying to answer the first part of
15
question one.
16
DR. PRZEPIORKA: Dr. Pazdur?
17
DR. PAZDUR: I would just like to
comment
18
that when we talk about response rates, remember
19
that the "other" drug that you mentioned was a
20
single agent that produced that 10 percent response
21
rate. We are talking about a
combination therapy
22
and, therefore, one has to take a look at that
172
1
combination.
2
Also, I think it is very important that we
3
perhaps discuss this issue more about the complete
4
responses. Remember, 3 of the
proposed 11 complete
5
responses were surgically induced.
As far as my
6
recollection of the protocol, there was no uniform
7
statement about how surgery was going to be
8
applied. This is really a very
down-the-line
9
analysis. There is a great deal
of subjective
10
bias. We all know who are
surgical candidates and
11 who
are not surgical candidates.
12
To the patients, I fully understand the
13
importance of complete responses and whether they
14 get
it by surgery plus chemotherapy or chemotherapy
15
alone probably may not matter to them.
What we are
16
addressing here though is a drug effect, and I
17
think it is important that we take a look really at
18
those surgically induced complete responses really
19 as
partial responses, if they were in fact, that
20
would then render them disease-free by surgery. I
21
think that would be a more appropriate way of
22
really suggesting this entire issue.
173
1
But this whole idea of surgery intervening
2
here--granted, it is very important--there is a
3
higher degree of subjectivity and unless that is
4
handled in a prospective manner on both arms of the
5
study it is really hard to ascertain how many
6
complete responses, especially when people are
7
following these patients out for prolonged periods
8 of
time--the symmetry of follow-up has to be
9
similar.
10
DR. PRZEPIORKA: Dr. Hwu?
11
DR. HWU: Regarding the response
rates to
12 the
single agent in the other Phase 3 trial, we
13
have to remember that although the response rate is
14
similar to this study, in that study it allowed 20
15
percent of the patients with brain metastases and
16 on
the DTIC arm all the 20 patients who had brain
17
metastases did not respond as compared to the 5
18
percent response. So, you have to
discount those
19 20
patients in that study.
20
DR. PRZEPIORKA: Thank you. If there are
21 no
other burning issues I would like to call the
22
question. Dr. Lopez?
174
1
DR. GRILLO-LOPEZ:
Grillo-Lopez. At the
2 end
of the session today we really have to address
3
question number five which, regardless of all of
4 the
above, is should Genasense be approved and made
5
available to the patients who need it?
That
6
relates to what Dr. Pazdur and Dr. Temple just
7
said. We need to give a
recommendation on whether
8 or
not there is an effect and if that effect is
9
important enough to merit approval of this agent,
10 and
that question is not asked so I would ask that
11 we
add that as question number five.
12
DR. PAZDUR: That is patient
access and I
13
think that is a different question.
There are
14
obviously access mechanisms available through
15
expanded access programs. We are
asking basically
16
about issues here that are defined in our
17
questions. If you would like to
discuss that at
18 the
end, please feel free to do so.
19
DR. PRZEPIORKA: Dr. Temple, do
you have
20 any
brief comments before we take a vote?
21
DR. TEMPLE: I just have one
thing. Maybe
22 you
will find it distracting. There is some
sense
175
1
that there is a small fraction of the population
2
that has a very special response and maybe, indeed,
3
that is true. But in the two
figures that we have
4
seen that look at that, namely progression-free
5
survival and survival itself, the curves at about
6 700
days are right on top of each other. In
fact,
7 for
progression-free survival Genasense is slightly
8
below. So, maybe the continued
data will show that
9
there is an excess of long-term survivors but at
10
least in the data we have seen so far it is very
11
hard to discern this hyper-responder group. I
12
don't know whether that is lack of maturity of the
13
data and when the last 10 percent of the people are
14
looked at something will turn up but, at least in
15
those figures, there is no hint of that and I just
16
wondered what everybody thinks about that in light
17 of
the possibility that there might be some people
18 who
get particularly good responses.
19
DR. PAZDUR: I think it is also
important
20
that people are cognizant, when they talk about
21
these responses, these complete responses, that the
22 N
in the treatment arm is quite high. We
are
176
1
talking about, whether one wants to say 8
2
responses, 10 responses, how many patients were in
3
that arm.
4
DR. PRZEPIORKA: So the survival issue
5
actually falls under question two I think and we
6
will discuss that in just a few moments.
Dr.
7
Cheson, you had some other comments?
8
DR. CHESON: Just one comment
about that.
9
Didn't they stop collecting survivor data at a
10
certain point for these curves and, therefore, we
11
don't know if they were censored--what?--at two
12
years or something and we don't know what goes on
13
beyond that.
14
DR. TEMPLE: That is what I am saying. As
15 far
as the data that we have been presented, you
16
don't see that tail on the curve looking different.
17 In
fact, they are right on top of each other.
18
Maybe with the final values on everybody you will
19 see
something but I don't see that yet, even though
20
there are obviously some people who had good
21
responses to either the drug or the combination.
22
DR. PRZEPIORKA: Let's go ahead
with the
177
1
vote and we are going to simply start at one end of
2 the
table and go around. Dr. Grillo-Lopez
and Dr.
3 Wen
Jen-Hwu are not voting members but everyone
4
else should give a yes, no or abstain.
5
Question 1A would be does the committee
6
believe that the observed differences in response
7
rate represent a real effect of Genasense when
8
added to DTIC? Dr. Bukowski, we
will start with
9
you.
10
DR. BUKOWSKI: No.
11 DR. BISHOP: Yes.
12
DR. PRZEPIORKA: Dr. Taylor?
13
DR. TAYLOR: No.
14
DR. REAMAN: Yes.
15
DR. REDMAN: Yes.
16
DR. PRZEPIORKA: Yes.
17
DR. RODRIGUEZ: Yes.
18
DR. DOROSHOW: Yes.
19
DR. CHESON: Yes.
20
DR. GEORGE: Yes.
21
MS. HAYLOCK: Yes.
22
DR. CARPENTER: Yes.
178
1
DR. D'AGOSTINO: No.
2
DR. MORTIMER: No.
3
DR. HUSSAIN: No.
4
MR. MCDONOUGH: Yes.
5
DR. GRILLO-LOPEZ: I am a
non-voting
6
member but I would vote yes if I were allowed to.
7 [Laughter]
8
So, the end of the vote says 11 yes and 5
9
no. Question 1B would be does the
committee
10
believe that the observed difference in
11
progression-free survival represents a real effect
12 of
Genasense when added to DTIC? We will
start
13
with Mr. McDonough and go the other way.
14
MR. MCDONOUGH: Yes.
15
DR. HUSSAIN: No.
16
DR. MORTIMER: No.
17
DR. D'AGOSTINO: No.
18
DR. CARPENTER: No.
19
MS. HAYLOCK: Yes.
20
DR. GEORGE: No.
21
DR. CHESON: Yes.
22
DR. DOROSHOW: No.
179
1
DR. RODRIGUEZ: No.
2
DR. PRZEPIORKA: No.
3
DR. REDMAN: Yes.
4
DR. REAMAN: No.
5
DR. TAYLOR: No.
6
DR. BISHOP: No.
7
DR. BUKOWSKI: No.
8
DR. PRZEPIORKA: The final vote is
6 yes
9 and
10 no. Let's move on to question
two. Do the
10
results of the study, in particular the difference
11 in
response rate and/or progression-free survival
12 for
the combination of Genasense and DTIC versus
13
DTIC alone, in the absence of a survival
14
improvement, provide substantial evidence of
15
effectiveness that outweighs the increased toxicity
16 of
administering the Genasense for the treatment of
17
patients with metastatic melanoma who have not
18
received prior chemotherapy?
19
While the members of the committee are
20
thinking about comments, I personally have two.
21 One
is that I know the folks at the FDA have seen
22 me
say, "yes, I'm a pro PFS kind of person" with
180
1 the
exception of when the experiment is not done
2
very critically. So,
progression-free survival I
3
think has to be considered a valid endpoint in
4
melanoma for which there is no drug that shows a
5
benefit for survival. There is no
question about
6
that.
7
The other issue has to do with the
8
administration. As was pointed
out, this is a drug
9
added to another drug and Genasense is administered
10 by
continuous infusion requiring a pump and a
11
catheter and is not given as a pill.
I think that
12
actually also weighs with regard to what I was
13
thinking.
14
I have just been handed a recount.
On
15
question 1B the recount is four yes and 12 no.
16
Thank you to the folks who went through the tape
17 and
listened to everyone once again. Other
18
comments on question two? Dr.
D'Agostino?
19
DR. D'AGOSTINO: I think we do, in
20
responding to question two, have to remember what
21 the
objective of the study was. The
objective of
22 the
study was to have a primary outcome of survival
181
1 and
some secondary outcomes, of which two are
2
mentioned here. The survival was
not significant
3 and
I am concerned or confused about where the
4
separation comes from. Maybe
later data will show
5 us
that but it is sort of beyond the study time
6
period and heaven knows what other things were
7
going on. So, again, to focus it,
we did have
8
survival as the primary outcome.
It wasn't
9
significant and the secondary outcomes weren't
10 obtained,
at least the progression wasn't obtained
11 in
the clearest fashion. So, I think we
have
12
concerns that the study didn't meet its objective.
13
DR. PRZEPIORKA: Dr. Lopez?
14
DR. GRILLO-LOPEZ: Grillo-Lopez; Lopez
is
15 my
mother's last name.
16
DR. PRZEPIORKA: I stand
corrected, thank
17
you.
18
DR. GRILLO-LOPEZ: Thank you. At the
19
December meeting of this committee we discussed
20
endpoints primarily in the setting of lung cancer.
21 But
as I recall, our recommendation to the FDA was
22 to
apply and utilize progression-free survival in
182
1
preference to overall survival in most settings.
2
There are some exceptions. So,
this protocol was
3
probably written four or five years ago and
4
discussed with the agency, and maybe at that time
5
overall survival was favored.
6
Now, those of you who are not familiar
7
with how primary endpoints are chosen should
8
understand that the sponsor meets with the agency
9 and
there are discussions around protocol design,
10 the
choice of endpoints and the statistical design
11 of
the study. And, it is not entirely up to
the
12
sponsor to choose the endpoints.
The agency, of
13
course, has a strong influence on what the primary
14 and
secondary endpoints are. I think it is
15
important, since it is an overriding concern for a
16
number of people here, the issue of not having met
17 the
primary endpoint--I think it is important to
18
know how the agency and the sponsor arrived at the
19
decision for that primary endpoint and whether or
20 not
that would have been the sponsor's first
21
choice.
22
DR. PRZEPIORKA: Dr. Temple?
183
1
DR. TEMPLE: Well, we have been
bringing
2 the
question of what the endpoint should be to
3
various deliberations of the advisory committee
4
for--I don't know, probably ten years; for a long
5
time. One of the problems that we
recognize is
6
that many trials have crossover and if there is
7
going to be crossover you have very little hope of
8
showing a survival effect. We
understand that.
9
That is a serious problem.
10
The other thing is that if death occurs
11
long after progression the numbers of people you
12
have to have in a trial to show a difference start
13 to
get huge even if you retain the whole benefit.
14 But
all of those conversations have reflected the
15
fact that disease-free survival has to be done
16
scrupulously, with great care, preferably in a
17
blinded study because it is subject to bias, and it
18 is
not just a simple matter of which do you like.
19 I
think that is what Rick said at the beginning,
20 and
that has always been part of the discussion
21
too. Whether people were
influenced by the
22
endpoints that we like or not, if somebody were
184
1
setting out to really do disease-free survival I
2
have to believe it would be done differently, and
3
that is part of the context too.
4
DR. GRILLO-LOPEZ: I think a lot
of us
5
don't like overall survival and that is the
6
discussion that we had in December.
Some of the
7
things that have to count against overall survival
8 as
an endpoint were mentioned by Dr. Pazdur
9
earlier. It is a biased endpoint
and those biases,
10 by
the way, were not mentioned by--
11
DR. TEMPLE: Why is survival a
biased
12
endpoint?
13
DR. GRILLO-LOPEZ: Let's go back
to the
14
December meeting. Survival as an
endpoint depends
15 on
an event, death. That event, if it
relates 100
16
percent exclusively to the disease, is useful. But
17
that is not reality. In the majority
of patients
18 it
doesn't relate 100 percent to the disease.
It
19
depends on complications of the disease or the
20
treatment. It depends on
co-morbidity, it depends
21 on
a variety--don't interrupt me; I am not
22
finished, Dr. Pazdur. Please turn
off your
185
1
microphone. Let me talk. You interrupted me once
2
before and that is enough. Okay?
3
The event is, in fact, something that can
4 be
manipulated. It can be manipulated
depending
5 on,
one, the supportive care the patient receives
6 or
does not receive. The patient may die
earlier
7 or
later because of that. That introduces a
bias.
8 The
event also depends on a death being certified
9 by
a physician who may or may not be the primary
10
physician, who may or may not know the patient and
11 the
natural history of his disease. So, if a
12
physician is seeing the patient for a first time at
13 the
deathbed and know the patient has cancer may
14 say
the cause of death, cancer. Maybe the
patient
15 had
an MI or pulmonary embolism. So, there
are
16
many ways in which overall survival is a biased
17
endpoint, which is why progression-free survival,
18
despite all of the problems that have been
19
mentioned here today about its measurement, is a
20
preferred endpoint because it is measurable.
21
DR. TEMPLE: There are
statisticians in
22 the
room. Most people wouldn't call bias in
any of
186
1
those things. That is an unusual
use of the term.
2
DR. PRZEPIORKA: If we could
continue with
3 the
discussion on question two which regards a
4
risk-benefit ratio, does the benefit, the small
5
benefit that has been seen in this particular study
6
outweigh the toxicities and the trouble with giving
7
everything by continuous infusion?
Dr. Carpenter?
8
DR. CARPENTER: I thought it was
worth
9
noting, in response to Dr. Temple's comments, that
10
long survival could confuse things because it
11
causes a death and could muddy the endpoint. Long
12
survival is not an issue in this study, at least
13
from what we have now. Since
there is no other
14
therapy which dependably prolongs survival in
15
melanoma, I think a crossover effect in this
16
population is extremely unlikely.
17
DR. PRZEPIORKA: Dr. D'Agostino?
18
DR. D'AGOSTINO: I just can't let
the
19
death be a biased endpoint. I am
sorry to eat up
20 the
time on the committee but I wish all studies
21 had
such a firm endpoint. The death is
all-cause
22 mortality; it is not cancer-related
mortality.
187
1
Right? So, we are not talking
about mistakes, and
2 I
hope that the investigators don't give
3
differential treatment to subjects depending on
4
what treatment they are on. So,
the biases that
5
might be generated by care I hope really are not an
6
issue.
7
DR. PRZEPIORKA: Any other
comments
8
regarding the toxicity and risk-benefit ratio? Dr.
9
George?
10
DR. GEORGE: I will pass.
11
DR. PRZEPIORKA: Dr. Hwu?
12
DR. HWU: We spent the last three
decades
13
trying to find standard care or better treatment
14 and
I believe all my colleagues in the field feel
15
that the only way to establish better treatment is
16
through a Phase 3 trial with an endpoint of
17
improved survival, not any other means because,
18
clearly, we have gone through this for years and
19 years
and improved response does not translate into
20
improved survival. The endpoint
has to be
21
survival, overall survival.
22
DR. PRZEPIORKA: Dr. Redman?
188
1 DR. REDMAN: Just for my clarification
2
because I really need things simplified, question
3 one
that I answered already is basically saying is
4
there a difference and do you believe the
5
difference is real. Question two
is asking us is
6 it
of clinical benefit.
7
DR. PAZDUR: That is the approval
8
question.
9
DR. PRZEPIORKA: Other
comments? If not,
10 I
will call the question. Do the results
of this
11
study, in particular differences in response rate
12
and/or progression-free survival for the
13
combination of Genasense plus DTIC versus DTIC
14
alone, in the absence of a survival improvement,
15
provide substantial evidence of effectiveness that
16 outweighs
the increased toxicity of administering
17
Genasense for the treatment of patients with
18
metastatic melanoma who have not received prior
19
chemotherapy? We will start with
Dr. Bukowski,
20
please.
21
DR. BUKOWSKI: No.
22
DR. BISHOP: No.
189
1
DR. TAYLOR: No.
2
DR. REAMAN: No.
3
DR. REDMAN: No.
4
DR. PRZEPIORKA: No.
5 DR. RODRIGUEZ: No.
6
DR. DOROSHOW: No.
7
DR. CHESON: Yes.
8
DR. GEORGE: No.
9
MS. HAYLOCK: Yes.
10
DR. CARPENTER: No.
11
DR. D'AGOSTINO: No.
12 DR. MORTIMER: No.
13
DR. HUSSAIN: No.
14
MR. MCDONOUGH: Yes.
15
DR. PRZEPIORKA: The final vote
then is
16
three yes and 13 no. The third
question has a
17
rather lengthy prologue. For
regular approval of a
18
drug for metastatic melanoma, the FDA has
19
considered an improvement in survival and/or
20
disease symptoms to constitute clinical benefit.
21
However, in the December ODAC discussion
22
considerable interest was expressed in
190
1
progression-free survival as an endpoint in some
2
settings, particularly where crossover to other
3
treatment could obscure a potential survival
4
benefit. In the metastatic
melanoma setting, do
5 you
believe that a progression-free survival
6
benefit of some magnitude represents clinical
7
benefit that could support regular drug approval,
8
even in the absence of an effect on survival?
9
We have initiated some discussion and I
10
will just throw my two cents in here and say
11
absolutely, in a disease where there is no drug
12
that confers a survival benefit having a
13
progression-free survival, getting patients off
14
chemotherapy for some period of time or at least
15
away from the stigma of having active disease is a
16
clinical benefit. Any other
comments from the
17
committee? Dr. George?
18
DR. GEORGE: Just a comment I made
19
actually at the last meeting when we discussed this
20 has
to do with the crossover effect issue that
21
people seem to obsess about quite a bit.
The real
22
point about that is that if there is something that
191
1
happens later that affects the outcome, then you
2
still can look at survival. That
is, there still
3 is
an answer. The answer may not be what
you
4
wanted to answer, that is, did this therapy prolong
5
survival if I didn't give anything else later or if
6 I
absolutely controlled everything precisely the
7
same way beyond this point? But
is the real-world
8
answer that in the current setting with available
9 therapies
that are so-called salvage therapies
10
sometimes and other things, it may not work with
11
respect to survival or it may work but the answer
12 is
still a good one for that therapy.
Having said
13
that, I still think that progression-free survival,
14
done properly, is in fact a very good way to do it.
15
DR. PRZEPIORKA: Dr. Carpenter?
16
DR. CARPENTER: I just second
that.
17
DR. PRZEPIORKA: Dr. Grillo-Lopez?
18
DR. GRILLO-LOPEZ: It is important
to
19
consider that for the majority of agents that come
20
before the FDA for approval the submission package
21
does not include data as to their optimal use,
22
perhaps the use with a combination therapy that may
192
1
have the potential of prolonging survival. Usually
2
this is the early data. It is the
first studies
3
done with an agent and you maybe will see evidence
4 of
clinical activity but not necessarily the
5
optimal use within the best possible combination of
6
that agent. There are many
examples of that.
7
I will give you rituxan, a product for
8
which I was responsible for clinical development.
9
When we presented the data to the agency we did not
10
have the optimal use of that agent that would
11
prolong overall survival. In
fact, that happened
12
only five to six years after the fact when the
13
combination with CHOP has shown that it can
14
increase the cure rate in patients with diffuse
15
lymphoma.
16
So, again, we have to be careful because
17
that is another problem with overall survival as an
18
endpoint. You seldom receive at
the
19
beginning--you, the agency, seldom receive at the
20
beginning the optimal use of the agent, and I think
21 you
have to be very careful and look for clinical
22
activity. If it has clinical
activity, then it
193
1
should be approved and it should go to the medical
2
community that really has the responsibility for
3
finding what the eventual optimal use in
4
combination, and so on, is for that agent.
5
DR. PRZEPIORKA: Dr. D'Agostino?
6
DR. D'AGOSTINO: Is it a quality
of life
7
issue that you are suggesting by using this
8
variable that the individual removes a stigma?
9
DR. PAZDUR: let me just jump in
here. Do
10 I
have permission to speak?
11
DR. PRZEPIORKA: Yes, sir.
12
DR. PAZDUR: Thank you. The issue here is
13
that we really brought this to the committee
14
because we really wanted to illustrate problems of
15
time to progression or progression-free survival.
16 In
order for this to have rigor it has to be
17
adequately measured and prospectively defined. The
18
points that I was trying to get across that I wrote
19
last night and read to you is that this is really
20
almost a harder endpoint to do correctly. It
21
requires robustness. It probably
requires that the
22
pharmaceutical sponsors actually meet with their
194
1
investigators and emphasize to them how to handle
2
missing data. The symmetry of
assessments have to
3 be
there. It actually is a much more
difficult
4
endpoint to assess.
5
Now, getting back to Dr. D'Agostino's
6
question, I think one of the fundamental issues
7
that you have to answer, and here again it comes
8
back to question number four, which is almost an
9
unanswerable question because it is in the eyes of
10 the
beholder--what is the magnitude? What is
the
11
benefit of delaying progression of a disease?
12
Here, again, in any analysis of survival with a
13
conventional toxicity profile, we have really not
14
answered that question if it was statistically
15
significant with an acceptable toxicity profile.
16 But
when you are dealing with a progression
17
endpoint, I think one has to ask oneself what is
18 the
benefit in light of the toxicity, even if the
19
toxicity is what one would encounter in a standard
20
chemotherapy drug.
21
The other issue that we have been
22
discussing with sponsors as we move away and we
195
1
have to ask ourselves why we should move away in
2
individual disease, and Bob brought this up, is
3
whether it is a problem with crossover.
Is the
4
disease of such sufficient natural history that is
5 so
long that a survival endpoint might not make
6
sense to bring up? Is the trial
so big that it is
7
unmanageable to do? Why does one
want to
8
substitute PFS for survival? That
may be an
9
individual disease setting that that needs to be
10
discussed, and that is why we are approaching these
11
disease by disease rather than just making a
12
uniform policy that we will no longer look at
13
survival; we will look at progression-free
14
survival.
15
The other issues that we have discussed
16 with sponsors is that we really like the
studies to
17 be
powered at least for survival, not that that
18
would necessarily be an approval endpoint, but it
19 is
something that I think we have to look at
20
eventually. We could approve a drug,
for example,
21 on
progression-free survival but if we never power
22 the
study for survival we will never know whether
196
1 any
of our treatments have a survival advantage and
2
that would really put medical oncology behind
3
significantly.
4
The other issue, finally, is power on
5
trials. To power a trial requires
a degree of
6
guesstimation and frequently we have seen trials
7 that
come to this committee as under-powered
8
trials. At least if we power for
survival, one
9
would hope that a progression-free survival would
10 be
adequately powered even with the uncertainties
11
that exist there.
12
DR. PRZEPIORKA: Dr. Redman?
13
DR. REDMAN: I agree that
progression-free
14
survival is probably important and I think one of
15 the
problems is the p value. If someone says
I am
16
going to power a trial to prove that for patients
17
getting drug X the progression-free interval is
18
three weeks greater and they had a p value with six
19
zeroes in front of it, the question is, no matter
20 how
rigorously it was done, how clinically relevant
21
that is. I guess it comes down to
the point, and
22 it
is not very scientific, that you will know it
197
1
when you see it.
2
DR. PRZEPIORKA: Dr. Temple?
3
DR. TEMPLE: A couple of other
points
4
while we are discussing this, there has never been
5 any
question that if someone had data on time to
6
symptomatic progression that would be a clinically
7
meaningful endpoint. Despite our
saying that at a
8
hundred end-of-Phase 2 conferences we have been
9
very unsuccessful at getting anybody to look at
10
that. I just want to make the
advert that even
11
after someone progresses radiologically you could
12
still measure time to symptomatic progression,
13
especially if there isn't anything very good to
14
transfer the patient to. So, that
is one pitch.
15
The second this is sort of a practical
16
matter. When you calculate the
increase in sample
17 size
that is needed to show survival, even if the
18
effect on survival was the same as the effect on
19
time to progression, if death occurs considerably
20
after progression the effect size gets depressingly
21
small. So, if you had a hazard
ratio of 0.8 at 10
22
months and survival goes to 20 months that same
198
1
difference becomes a hazard ratio of 0.9 and the
2
sample size implications become quite daunting.
3
That is a practical concern but it could mean that
4
trials in that setting would have to be just
5
enormous, and that is another reason we are
6
thinking about disease-free survival.
7
DR. PRZEPIORKA: Just to come back
to a
8
question that Dr. D'Agostino asked me earlier, you
9
raised the issue of symptomatic relapse and I still
10
have great concerns that depression and anxiety are
11
truly symptoms that we wish to address.
Dr.
12
Carpenter?
13
DR. CARPENTER: I think how much
one is
14
willing to accept a progression-free survival
15
endpoint is going to be inevitably tied to question
16
four but a couple of simple examples help to modify
17 the
way one might think about it. In this
18
application that we are discussing the issues were
19 all
with a possible increase in progression-free
20
survival on the order of magnitude of a month or
21
less, no matter which projection you look at. If
22 you
were talking about something in the 3-6 month
199
1
interval I would be surprised if the tenor of the
2
discussions was not different and if the difference
3 in
survival, even if it was small, would not become
4
secondary. The more we get into
drugs that act by
5
biological mechanisms that may not shrink tumors
6 but
which might stop growth so you may get long
7
periods and if you get relief of symptoms and
8
prolonged freedom from progression, I think it
9
would be an unusual person who won't think that is
10 a
benefit.
11
The question in this particular
12
application was whether they have really met some
13
kind of endpoint that would be satisfactory. Could
14 one
accept unequivocally that they have met that or
15
not, and the votes are there.
16
DR. PRZEPIORKA: Ms. Haylock?
17
MS. HAYLOCK: Let's see, all the
numbers I
18 think
kind of obscure the reality of what melanoma
19
patients face and I think of all the kinds of
20
cancers, the dying process in melanoma is sometimes
21
long and drawn out and fairly awful.
So, I think
22
that symptomatic progression is important just in
200
1
terms of the things that people do go through if
2
their treatment fails overall.
3
So, I think the cure versus control issue
4 we
are looking at in this particular kind of
5
cancer, like a lot of cancers, is more of a chronic
6
disease entity and how do we control those chronic
7
symptoms for longer periods of time and give people
8
quality for whatever time they have left--I think
9
that is sort of lost in all the numbers,
10
particularly lost when people just look at death as
11 the
sentinel event in this.
12
DR. PRZEPIORKA: If there are no
other
13
questions I will ask for a vote.
Question number
14
three, in metastatic melanoma, do you believe that
15 a
progression-free survival benefit of some
16
magnitude represents clinical benefit that could
17
support regular drug approval, even in the absence
18 of
an effect on survival? Mr. McDonough?
19
MR. MCDONOUGH: Yes.
20
DR. HUSSAIN: Yes.
21
DR. MORTIMER: Yes.
22
DR. D'AGOSTINO: Yes.
201
1 DR. CARPENTER: Yes.
2
MS. HAYLOCK: Yes.
3
DR. GEORGE: Yes.
4
DR. CHESON: Yes.
5
DR. DOROSHOW: Yes.
6
DR. RODRIGUEZ: Yes.
7
DR. PRZEPIORKA: Yes.
8
DR. REDMAN: Yes.
9
DR. REAMAN: Yes.
10
DR. TAYLOR: Yes.
11
DR. BISHOP: Yes.
12
DR. BUKOWSKI: Yes.
13
DR. PRZEPIORKA: It is unanimous,
yes.
14 The
last question for discussion, which we have had
15 a
tremendous amount about is, if yes, please
16
discuss what magnitude of improvement in this
17
endpoint would be required to demonstrate clinical
18
benefit and whether this would depend on the
19 toxicity
of the treatment.
20
I will just start by saying not just
21
toxicity of the treatment but the way the drug is
22
administered, and in this situation where the drug
202
1 was
administered by continuous infusion for a
2
patient population who had no other alternative,
3
like many diabetics who are on a fanny-pack right
4
now, I don't think the patients would mind having
5 the
fanny-pack for the rest of their life if it
6
meant they would actually get a clinical benefit
7
from it. So, for this particular
setting how the
8
drug is administered is less of an issue because of
9 the
background.
10
Other comments regarding this question
11
from the committee? Hearing none,
Dr. Temple and
12 Dr.
Pazdur, do you have any other questions you
13
need advice on from us?
14
DR. PAZDUR: No.
15
DR. PRZEPIORKA: Thank you. I call this
16
meeting adjourned then. We will
meet here promptly
17 at
12:45 to begin the second session. Thank
you.
18
[Whereupon, the proceedings were recessed
19 for
lunch, to reconvene at 12:45 p.m.]
203
1
A F T E R N O N P R O C E E D I N
G S
2
DR. PRZEPIORKA: In the interest
of time,
3 we
will start the meeting and we will have a few
4
people in and out during the course of the day, and
5 I
apologize but we do want to stay on time as much
6 as
possible.
7
This afternoon we will be discussing RSR13
8 and
we want to start with a conflict of interest
9
statement. I understand there are
no conflicts of
10
interest for the group for this afternoon. Please
11
refer to this morning's statement if you want more
12
information.
13
Because we have moved around a bit and
14
there are new individuals who have joined us for
15
this particular meeting, I would like to go ahead
16 and
allow the committee to introduce themselves
17
once again and if we could start with Ms. Portis.
18
MS. COMPAGNI-PORTIS: Natalie
19
Compagni-Portis. I am a patient
representative.
20
DR. MORTIMER: Joanne Mortimer,
medical
21
oncology, Eastern Virginia Medical School.
22
DR. HUSSAIN: Maha Hussain,
medical
204
1 oncology,
University of Michigan.
2
DR. D'AGOSTINO: Ralph D'Agostino,
Boston
3
University, biostatistician.
4
DR. BUKOWSKI: Ronald Bukowski,
medical
5
oncologist, Cleveland Clinic.
6
DR. BUCKNER: Jan Buckner, medical
7
oncology, Mayo Clinic, Rochester, Minnesota.
8
DR. MARTINO: Silvana Martino,
medical
9
oncology, the John Wayne Cancer Institute.
10
DR. TAYLOR: Sarah Taylor, medical
11
oncology, Palliative Care, University of Kansas.
12
DR. REAMAN: Gregory Reaman,
pediatric
13
oncologist, George Washington University and the
14
Children's Hospital.
15
DR. REDMAN: Bruce Redman, medical
16
oncologist, University of Michigan.
17
MS. CLIFFORD: Johanna Clifford,
FDA,
18
executive secretary to this meeting.
19
DR. PRZEPIORKA: Donna Przepiorka,
20
hematology, University of Tennessee, Memphis.
21
DR. RODRIGUEZ: Maria Rodriguez,
medical
22
oncologist, M.D. Anderson Cancer Center.
205
1
DR. DOROSHOW: Jim Doroshow,
Division of
2
Cancer Treatment and Diagnosis, NCI.
3
DR. GEORGE: Stephen George, Duke
4
University.
5
MS. HAYLOCK: Pamela Haylock,
oncology
6
nurse.
7
DR. CARPENTER: John Carpenter,
medical
8
oncologist, University of Alabama at Birmingham.
9
DR. RIDENHOUR: Kevin Ridenhour,
medical
10
reviewer, FDA.
11
DR. SRIDHARA: Rajeshwari
Sridhara,
12
statistical reviewer, FDA.
13
DR. DAGHER: Ramzi Dagher, medical
team
14
leader, FDA.
15
DR. WILLIAMS: Grant Williams,
Deputy
16
Director, Oncology Drugs.
17
DR. PAZDUR: Richard Pazdur,
Director,
18
Oncology Drugs.
19
DR. TEMPLE: Bob Temple, Office
Director.
20
DR. GRILLO-LOPEZ: Antonio
Grillo-Lopez,
21 Neoplastic
and Autoimmune Diseases Research
22
Institute.
206
1
DR. PRZEPIORKA: Thank you and
welcome to
2
all. I just again want to remind
everyone in the
3
room, as well as on the committee, that this is a
4
committee that serves as consultants to the FDA.
5 We
are not employed by the FDA or the U.S.
6
government. We do not make any
decisions here; we
7
simply provide advice to the FDA.
8 We will start the presentations
this
9
afternoon with Dr. Pablo Cagnoni, from Allos, to
10
introduce the topic.
11 Sponsor Presentation
12 Introduction
13
DR. CAGNONI: Good afternoon, Dr.
14
Przepiorka, ladies and gentlemen.
15
[Slide]
16
My name is Pablo Cagnoni and I am
17
representing Allos Therapeutics today for this
18
presentation to the Oncologic Drugs Advisory
19
Committee for the new drug application for RSR13 as
20 an
adjunct to whole brain radiation therapy for
21
patients with breast cancer and brain metastases.
22
[Slide]
207
1
Our agenda for today is shown here.
After
2 a
brief introduction Dr. John Suh will provide an
3
overview of brain metastasis.
This will be
4
followed by Dr. Brian Kavanaugh who will provide a
5
review of the mechanism of action of RSR13, early
6
preclinical and clinical data. I
will then
7
summarize the efficacy and safety data with our
8
compound and we will have some concluding remarks
9 by
Dr. Paul Bunn.
10
[Slide]
11
We have a number of experts today
12
available for the question and answer session: Dr.
13
Paul Bunn, Director of the University of Colorado
14
Cancer Center; Dr. Walter Curran, Group Chairman of
15 the
Radiation Therapy Oncology Group; Dr. Anthony
16
Elias, Director of the Breast Cancer Program at the
17
University of Colorado.
18
[Slide]
19
Dr. Henry Friedman, Director of the Brain
20
Tumor Center at Duke University Medical Center; Dr.
21
Marc Gastonguay, clinical pharmacologist who
22
performed the clinical pharmacokinetic analysis and
208
1
population pharmacokinetic analysis for RSR13; Dr.
2
Charles Scott, biostatistician, former statistician
3
from RTOG who conducted the analysis of our RT-08
4 and
served as a design analysis consultant for
5
RT-09; Dr. Baldassarre Stea, Chairman, Radiation
6
Oncology at the University of Arizona, who is a
7 lead
enroller in study RT-09.
8
[Slide]
9
In addition, we have a number of experts
10
from Allos Therapeutics that will be available to
11
answer questions as well.
12
[Slide]
13
We need to acknowledge today that brain
14
metastases in patients with breast cancer represent
15 an
unmet medical need. This complication
afflicts
16
tens of thousands of patients a year in the U.S.
17
alone. It carries a very high
morbidity and nearly
18
uniform mortality. This field has
been
19
characterized for the last 25 years by lack of
20
progress in terms of improving the survival of
21
these patients. The data that we
will review for
22 you
today demonstrates that RSR13 improves the
209
1
survival of patients with breast cancer and brain
2
metastases; increases the response rate in the
3
brain in these patients; and has an excellent
4 safety
profile in this population.
5
[Slide]
6
Our proposed indication for RSR13 is to be
7
administered as an adjunct to whole brain radiation
8
therapy for the treatment of brain metastases
9
originating from breast cancer.
Our proposed
10
dosage is RSR13 75-100 mg/kg/day IV over 30 minutes
11
with supplemental oxygen immediately prior to each
12 of
10 fractions of whole brain radiation therapy.
13
[Slide]
14
At this point, I would like to introduce
15 Dr.
John Suh. Dr. Suh is Clinical Director
of
16
Radiation Oncology and Director of the Gamma Knife
17
Radiosurgery Center from the Brain Tumor Institute
18 and
the Cleveland Clinic Foundation. Dr. Suh
was
19 the
study chair for our pivotal trial RT-09 and he
20 has
extensive experience with use of RSR13 in the
21
treatment of brain metastases.
22 Brain Metastases
210
1
DR. SUH: Good afternoon, ladies
and
2
gentlemen. It is a pleasure to be
here today to
3
talk about brain metastases. As a
clinician who
4
focuses his clinical and research efforts on brain
5
tumor patients, I have the opportunity to evaluate
6 and
treat a number of these patients. For
the past
7 ten
years I have been involved in a number of
8
clinical trials related to these patients and hope
9
that after today's discussion you will consider
10
changing the treatment paradigm for patients with
11
breast cancer who develop brain metastases.
12
[Slide]
13
In terms of the brain metastasis, its
14
incidence is on the rise. Every
year in the United
15
States approximately 170,000 Americans are
16
diagnosed with this condition. It
is estimated
17
that 20-40 percent of cancer patients will
18
eventually develop brain metastases.
The incidence
19 is
thought to be rising secondary to earlier
20
diagnosis of the cancer; better systemic therapy
21 for
extracranial disease; and better neuroimaging
22
techniques, the MRI scans.
211
1
[Slide]
2 In terms of breast cancer patients
with
3
brain metastases, up to 35,000 patients per year
4 are
diagnosed with this disease. It afflicts
5
younger patients. The median age
for our study was
6 53
years of age, and most of these patients are
7
quite functional as well.
Systemic agents have
8
provided benefit for extracranial disease.
9
Therefore, to control the brain becomes very
10
important. Current treatment
strategies have
11
provide limited benefit and, as a result, more
12
effective treatment options are needed.
13
[Slide]
14
This is an example of a an excellent
15
response from radiation therapy.
This is a picture
16 of
a CT scan of a patient with two very large brain
17
tumors in the frontal area, and after radiation
18
therapy you can see a dramatic response.
19
Unfortunately, this is a very untypical response
20
from radiation therapy and, as a result, we need
21
better therapies for these patients.
22
[Slide]
212
1
In terms of the current treatment
2
strategies for patients with brain metastases,
3
there are a number of treatment strategies
4
depending on the patient and their performance
5
status. Steroids have been shown
to increase
6
survival by approximately one month.
7
Anticonvulsant medication is used to prevent
8
seizures. Surgical resection has been
shown by
9
several randomized studies to improve survival for
10
patients with single metastases.
Stereotactic
11
radiosurgery has been shown by a recent trial to
12
improve survival for patients with a single lesion.
13
Chemotherapy has had limited use thus far. Whole
14
brain radiation therapy has been the gold standard
15 and
has been used for over 50 years for treatment
16 of
brain metastases.
17
[Slide]
18
In terms of the results with whole brain
19
radiation therapy, the mean survival is
20
approximately 4.5 months. it
improves and/or
21
stabilizes neurologic function in the majority of
22
these patients. The standard
dosing scheme
213
1
established by the RTOG is 30 Gy in 10 fractions.
2
There has been no benefit to altered fractionation
3
schemes.
4
[Slide]
5
This slide summarizes the lack of progress
6 over the past 20 years for patients with brain
7
metastasis. These are series from
the 1970s to the
8
1990s, looking at various fractionation schemes.
9 If
you look at the median survivals overall, they
10
range from about 3-5 months. Therefore, better
11
treatment is needed for these patients.
12
[Slide]
13
It is important when analyzing patients
14
with brain metastasis to have common prognostic
15
factors. RTOG performed a
recursive partitioning
16
analysis of 1200 patients enrolled in 3 consecutive
17
clinical trials from 1979 to 1993.
They came up
18
with 3 classes of patients. The
best class of
19
patients is Class I patients, with a KPS of 70 or
20
higher; primary controlled; age less than 65; and
21 no
extracranial metastasis, which comprised 20
22
percent of this database with median survival of
214
1 7.1
months.
2
For Class II patients, these are patients
3
with a KPS of at least 70 and any of the following,
4
controlled primary; extracranial metastases; age
5
greater than or equal to 65. This
comprises the
6
majority of the patients in this database; 65
7
percent survival of only 4.2 months.
8
For the Class III patients, these are
9
patients with a KPS less than 70; median survival
10 of
only 2.3 months, and resulting in poor survival
11 for
this group of patients. They are
typically
12
excluded from clinical trials.
13
[Slide]
14
If you focus on the results of whole brain
15
radiation therapy for patients with breast cancer,
16
these are some recent publications from the late
17 '90s to 2000, looking at 100 patients. You can see
18
here that their median survival has hovered between
19 4-6
months. The RTOG brain metastasis
database
20
that I alluded to, for 113 patients with brain
21
metastases, the median survival was 5.4 months.
22
This is a retrospective series from the
215
1
Cleveland Clinic of 116 patients.
When we looked
2 at
the one-year survival, it was only 17 percent
3 and
two-year survival was only 2 percent.
4
[Slide]
5
The recursive partitioning analysis
6
developed at the RTOG was consistent with the
7
control arm of the RT-009 study.
As you can see
8
here, for the Class I patients, 7.7 months versus
9 7.1
months, and for the Class II patients, 4.1
10
months versus 4.2 months, suggesting that this
11
database is reliable for comparing results.
12
[Slide]
13
In conclusion, brain metastases from
14
breast cancer are common. Current
treatment
15
strategies yield poor results.
Treatment options
16 are
available for extracranial metastases.
17
Therefore, it is paramount that we control the
18
disease within the brain to improve survival for
19
these patients, and there is a compelling need for
20
more effective treatment options.
21
[Slide]
22
At this point, I would like to introduce
216
1 Dr.
Brian Kavanaugh, who will talk about the
2
science of RSR13.
3 The Science of RSR13
4
DR. KAVANAUGH: Thank you,
John. It is an
5
honor to be here today. I have
been working with
6
RSR13 for ten years. I
participated in the
7
preclinical evaluation. I served
as the PI for the
8
Phase 1 study in cancer patients and I have
9
enrolled patients on both the Phase 2 and Phase 3
10
studies that you will be hearing about today.
11
[Slide]
12
In this section we will review several
13
topics, first of all, a brief refresher on tumor
14
hypoxia and its particular importance in
15
radiotherapy. We will explain how
and why RSR13
16 was
designed. We will explain how RSR13
improves
17
tumor oxygen delivery and, thus, radiosensitizes
18
solid tumors. And, we will share
some key
19
observations when the agent was first taken into
20 the
clinic.
21
[Slide]
22
Oxygen has long been recognized to be the
217
1
purest and most efficient radiosensitizer.
2
Ionizing radiation introduces free radicals which,
3 in
the presence of oxygen, are stabilized.
When
4
cancer cells are treated with radiotherapy in
5
oxygenated conditions the effect of radiation is
6
roughly tripled when compared with treatment with
7
radiation in hypoxic settings.
There are pockets
8 of
hypoxia or low pO-2 to varying extent in all
9
solid tumors. The reason this
exists is that
10
supply simply doesn't keep up with demand in
11
hyper-metabolic areas. It is
possible to measure
12 directly
in the clinic the degree of tumor hypoxia
13
present in certain solid tumors and in all cases
14
where this has been performed there is a direct
15
correlation between the extent of hypoxia and the
16
outcome after radiotherapy. Specifically,
the more
17
hypoxic the tumor is the lower the chance of
18
controlling with radiation.
19
I should just add one more point, that it
20 is
essential for the oxygen to be present at the
21
moment of radiation. The radiation-induced
free
22
radicals that are generated in the absence of
218
1
oxygen have a half-life of 10
-5 or 10-9 seconds and
2
with oxygen present this half-life is extended to
3 the
range of milliseconds. Nevertheless, it
is
4
important for oxygen to be present at the moment
5
that radiation is given.
6
[Slide]
7
To consider hypoxia in breast cancer in
8
particular, these data represent thousands of
9
individual point measurements of pO-2 within tumors
10 in
a cohort of breast cancer patients. On
the X
11
axis is the tissue oxygen pressure and on the Y
12
axis is the frequency with which a value and the
13
range shown on the X axis was observed.
14
You can see that fully 15 percent of the
15
measurements were less than 5 mmHg and this would
16 be
an extent of hypoxia expected to cause
17
substantial radioresistance. Now,
it is
18
technically very challenging to obtain pO-2
19
measurements clinically in tumors, and particularly
20
difficult in the brain. So, there
are far fewer
21 data
particularly with brain metastasis but what is
22
available would suggest that the rate of hypoxia is
219
1
probably even higher when tumors have spread to the
2
brain.
3 [Slide]
4
In the early 1980s Professor Don Abraham
5 and
the Nobel Laureate Max Perutz set out on a
6
mission to design agents which would have
7
therapeutic benefit by modifying the properties of
8
hemoglobin, and RSR13 is the product of their
9
collaboration.
10
As you can see here, RSR13 binds within
11 the
central water cavity of hemoglobin and exerts
12 an
effect on hemoglobin through a process called
13
allosteric modification. Under
the influence of
14
RSR13 hemoglobin is changed in its properties.
15
Specifically, the binding affinity between
16
hemoglobin and oxygen is reduced.
17
[Slide]
18
I will illustrate that for you in this
19
graph. You will recall that under
ordinary
20
conditions, represented here by the black curve,
21
there is an approximately sigmoidal relationship
22
between pO-2 in the bloodstream and the percent of
220
1
saturation of all available hemoglobin binding
2
sites. RSR13 has the property of
shifting this
3
curve right-ward. We can easily
quantify this
4
effect in terms of the p50. The
p50 is defined as
5 a pO-2 at which there is 50 percent
saturation of
6 all
available hemoglobin sites. We have
calculated
7 in
other studies that an increase in p50 of 10 mmHg
8 is
expected to have a major improvement on tumor
9
oxygen delivery and, thus, radiosensitization.
10
But before we leave this slide, let me
11
share one other particularly important point
12
regarding the reason why supplemental oxygen is
13
given to patients who receive RSR13.
At sea level
14 under
ordinary conditions you will recall that the
15
pO-2 of arterial blood is typically in the range of
16
90-100 mmHg. Under normal
conditions there would
17 be
expected to be 96-98 percent or so saturation of
18
hemoglobin binding sites. Adding
additional oxygen
19 in
that setting is unlikely to yield any noticeable
20
benefit because the blood is already carrying as
21
much oxygen as possible into the peripheral
22
circulation. Under the influence
of RSR13, in
221
1
order to exploit the agent to its maximal effect,
2 we
want there to be as high as possible saturation
3 of
blood leaving the lungs and entering the
4
peripheral circulation. That is
why we give
5
supplemental oxygen to achieve pO-2s in the range
6 of
120 or more so that blood leaving the lungs is
7
going to be at a very high level of oxygen
8
saturation.
9
[Slide]
10
There have been numerous clinical studies
11 to
establish both the proof of principle and the
12
establishment of the radiosensitizing effect of
13
this agent and I will share with you a couple of
14
examples.
15
In this situation, using a rodent mammary
16
carcinoma, the experimental endpoint was percent of
17
tumor oxygen pO-2 readings below 5 mmHg.
You can
18 see
in the yellow bar that under controlled
19
conditions this particular tumor is roughly 50
20
percent hypoxic. Oxygen has only
a modest effect,
21 and
I should add that in animals the reason for a
22
modest effect in oxygen in this kind of experiment
222
1 is
because they are anesthetized and there is a
2
certain amount of hyperventilation.
It is not
3
expected to have that much effect in humans. The
4
addition of RSR13 has an even stronger effect than
5
oxygen alone, and the combination of RSR13 and
6
supplemental oxygen essentially abolishes all
7
measurable tumor hypoxia. This
effect on tumor
8
oxygen levels translates directly into
9
radiosensitizing properties.
10
[Slide]
11
Again using a rodent model in the lab, the
12
experimental endpoint here is the clonogenic
13
survival fraction after in vivo exposure. With
14
RSR13 alone and oxygen, you can see that there is
15 no
appreciable effect on tumor cell surviving
16
fraction because the agent itself is not directly
17
cytotoxic. Radiation has, of
course, an expected
18
effect in terms of reducing tumor cell survival
19
fraction, but the combination of RSR13 and oxygen
20
will meaningfully sensitize cells to radiation and
21
have a pronounced additional radiosensitizing
22
effect.
223
1
This proof of principle and
2
radiosensitizing effect has demonstrated in
3
non-small cell lung cancers also and, in fact, for
4 all
solid tumors tested in the lab that RSR13 can
5
exert a radiosensitizing effect.
6
[Slide]
7
The first instance in which this agent was
8
taken into humans was in a study of healthy
9
volunteers. The targeted
pharmacodynamic endpoint
10 was
an increase of p50 of 10 mmHg which, as I have
11
already mentioned, is expected to have a meaningful
12
improvement in tumor oxygen delivery.
13
A Phase 1 study was conducted of 19
14
patients in which RSR13 was given in doses ranging
15
from 10 up to 100 mg/kg using a single intravenous
16
dose. The observation was an
increase in p50 of 10
17
mmHg achieved consistently at a dose of 100 mg/kg.
18
[Slide]
19
A few observations about the
20
pharmacokinetics of RSR13, its volume of
21
distribution is a vascular compartment.
Half the
22
drug is gone within red blood cells and the other
224
1
half is in plasma, most of it bound to plasma
2
proteins. The half-life in red
blood cells is 4.5
3
hours. The drug is partially
glucuronidated in the
4
liver and then both the parent compound and the
5
metabolites formed are excreted through the
6
kidneys.
7
[Slide]
8
The pharmacokinetic and pharmacodynamic
9
parameters analyzed in several studies have been
10
combined and the results are shown here.
In four
11
separate studies involving both the healthy
12
volunteers and a broad range of cancer patients,
13 the
pharmacokinetic parameter of mean red blood
14
cell concentration was assayed and directly
15 compared
with the mean p50 increase or
16
pharmacodynamic effect. The eight
data points on
17
this particular graph represent the averages of
18
those two groups of patients either receiving 75
19
mg/kg or 100 mg/kg in the four individual studies.
20
What you notice is a linear correlation
21
between these two parameters. On
the X axis again
22 is
the mean red blood cell concentration.
In order
225
1 to achieve our desired pharmacodynamic effect,
an
2
increase of 10 mmHg, we need to achieve in red
3
blood cells a concentration on the order of 480
4
mcg/mL.
5
[Slide]
6
Let me just summarize that tumor hypoxia
7 has
long been recognized to be a major cause of
8
radioresistance. RSR13 has the
properties of
9
reducing tumor hypoxia and increasing
10
radiosensitivity. The
pharmacodynamic effect of
11 the
agent is easily quantified by characterizing
12 the
increase in p50. There is a linear
correlation
13
between the drug concentration and the
14
pharmacodynamic effect. And,
RSR13, at a dose of
15 100
mg/kg, was selected for future study based on
16 its
ability to induce the desired p50 increase.
17
[Slide]
18
Now I will let Dr. Cagnoni present to you
19 the
clinical efficacy results.
20 Clinical Efficacy Results
21
DR. CAGNONI: Thank you, Dr.
Kavanaugh.
22
[Slide]
226
1
Today's presentation is a culmination of
2
almost ten years of clinical development of RSR13.
3
This was initiated with filing IND 48-171 in 1995.
4
This was followed by the human volunteer study that
5 Dr.
Kavanaugh described and, in turn, that was
6
followed by Phase 1 studies in combination with
7
radiation therapy. Our pivotal
study in patients
8
with brain metastases started enrollment in
9
February of 2000, completed enrollment in July 2002
10 and
the present NDA was submitted in December of
11
2002.
12
[Slide]
13
Before we describe the results of the
14
Phase 2 and Phase 3 studies, it is important to
15
understand how RSR13 is administered relative to
16
radiation in both studies. On
arrival to the
17
clinic oxygen and pulse oximetry for monitoring are
18
initiated. RSR13 is administered
through a central
19
venous access device over a 30-minute infusion in
20
both studies. Both studies
mandated that patients
21 be
radiated within 30 minutes of completing the
22
RSR13 infusion. After radiation
therapy was
227
1
administered patients were monitored as the oxygen
2 was
tapered, and they were released from the clinic
3
when oxygen saturation at room was acceptable. The
4
same process was repeated daily for 10 days.
5
[Slide]
6
Our Phase 2 study in patients with brain
7
metastases is study number RT-08.
It enrolled 69
8
patients. It was an open-label
study and 21 of the
9
patients in this study had breast cancer, 39 had
10
non-small cell lung cancer and 9 patients had other
11
tumor types. Patients were
enrolled at 17 sites in
12 the
U.S. and Canada. The primary endpoint of
the
13
study was survival. To use as a
comparison group
14 we
selected the RTOG brain metastasis database that
15 Dr.
Suh summarized for you earlier.
16
[Slide]
17
When we compared the results of the RT-08
18
Class II patients with the RTOG database Class II
19
patients, we see the following results:
In yellow
20 are
the RSR13 patients with a median survival of
21 6.4
months and in red is the median survival of
22
4.11 with the patients in the RTOG brain metastasis
228
1
database.
2
[Slide]
3
We then compared these two groups by tumor
4
type within the Class II patients, and in breast
5
cancer of the RTOG database there was a median
6
survival of 5.4 months and in the RSR13-treated
7
patients the median survival was 9.7 months. In
8 the
lung cancer population the survival was 3.9 and
9 6.4
months respectively.
10
[Slide]
11
As a result of this study a pivotal trial
12 was
initiated, study number RT-09. This was
a
13
Phase 3 randomized, open-label, comparative study
14 of
standard whole brain radiation therapy with
15
supplemental oxygen, with or without RSR13, in
16
patients with brain metastases.
The study chairs
17
were Dr. John Suh, from the Cleveland Clinic, and
18 Dr.
Edward Shaw from Lake Forest University.
19
[Slide]
20
The key eligibility criteria for RT-09 are
21
summarized here. Patients had to
have a KPS of at
22
least 70. In other words, Class
II patients were
229
1
excluded. The excluded
histologies were small-cell
2
lung cancer, non-Hodgkin's lymphoma and germ cell
3
cancer. No prior therapy for
brain metastases was
4
allowed, with the exception of partial resection.
5 In
other words, patients had to have measurable
6
disease after resection. All
patients had to have
7
adequate hematologic, renal, hepatic and pulmonary
8
function, including resting and exercise oxygen
9
saturation of at least 90 percent on room air.
10
[Slide]
11
This was a 1:1 randomization. It
was an
12
open-label study. All patients
received standard
13
whole brain radiation therapy, 3 Gy fractions for
14 10
days for a total of 30 Gy. Both arms
received
15
supplemental oxygen and patients were randomized to
16
receive or not RSR13. At the time
of randomization
17
patients were stratified using RPA class and tumor
18
type.
19
The primary endpoint of RT-09 was
20
survival. The study had 85
percent power to detect
21 a
difference in all patients and 75 percent power
22 to
detect a difference in the lung/breast
230
1
co-primary population. These are
the only two
2
populations for which the alpha spending and the
3
log-rank test was calculated.
4
[Slide]
5
RT-09 was amended three times, generating
6
four protocol versions. The key
amendment in the
7
study is amendment two. Amendment
two took place
8
between versions two and three.
At the time of the
9
amendment 222 patients had been enrolled in the
10
study. The key components of the
amendment were to
11
expand the sample size up to 538 patients; to
12
define the lung/breast co-primary population as a
13
co-primary population for analysis; and it expanded
14 the
dosing adjustment guideline of RSR13 for
15
patients receiving antihypertensive medications,
16
including also weight and gender.
This amendment
17 was
discussed with the FDA at the time and
18 concurrence
was reached on the approvability of
19
this co-primary population.
20
[Slide]
21
The dosing adjustment guideline is
22
summarized here. Using the weight
cutoff of 70 kg
231
1 for
women and 95 kg for men, the study divided
2
patients in high weight/low weight categories.
3
According to the guideline, high weight patients
4
were to receive an initial dose of RSR13 of 75
5 mg/kg
and low weight patients were to receive a
6
dose of RSR13 of 100 mg/kg.
7
[Slide]
8
For the primary endpoint of survival we
9
assumed that 20 percent of the patients would be
10 RPA
Class I. We expected a median survival
time in
11 the
control arm of 4.57 months and a 35 percent
12
improvement over this would have been a median
13
survival of 6.17 months in the RSR13 arm. The
14
analysis of the study was determined by a number of
15 events,
with a minimum follow-up of 6 months and
16
minimum number of events or 402 patients had to
17
occur in all patients and the minimum number of
18
events of 308 had to occur in the lung
19
cancer/breast cancer co-primary population.
20
[Slide]
21
The analysis of survival following the
22
statistical analysis plan, which was completed
232
1
prior to the completion of enrollment, defined that
2 the
primary method for survival analysis would be
3 an
unadjusted log-rank. The primary
population for
4
analysis of survival would be comprised of the
5
eligible patients. For the
co-primary population
6 of
lung and breast cancer patients a modified
7
Bonferroni adjustment was described
Both the
8
protocol and the SAP specified the Cox multiple
9
regression analysis would be conducted.
10
[Slide]
11
The benefits of this type of analysis are
12
summarized here. Adjusted
analyses, such as Cox or
13
stratified log-rank, provide the most accurate
14
treatment estimate in heterogeneous populations.
15 As
we will see in the presentation, the population
16 of
patients in RT-09 was clearly very
17
heterogeneous. It is important to
remember that
18
omitting strong covariates can reduce the power of
19 the
study to detect treatment effects.
20
[Slide]
21
To this effect, prespecification of the
22 Cox
model was performed in the protocol and
233
1
expanded in the statistical analysis plan. Seven
2
covariates, in yellow, were specified in the
3
protocol and were derived from the literature. In
4
addition to this, ten more covariates were added in
5 the
statistical analysis plan. The top six
in
6
yellow are derived from the literature as well.
7 The
bottom four were specific to the study to take
8
into account the mechanism of action of RSR13 and,
9 i
the case of the weight category to take into
10
account the dosing adjustment guideline.
11
[Slide]
12
RT-09 had five secondary endpoints.
The
13
objective of RSR13 is to improve local therapy in
14 the
brain, therefore, the most important secondary
15
endpoint in the study is response rate in the
16
brain. Other secondary endpoints
were time to
17
radiographic tumor progression in the brain and to
18
clinical tumor progression in the brain, cause of
19
death and quality of life.
20
[Slide]
21
For the radiologic evaluation the
22
following was mandated by the protocol, all
234
1
patients had to have a CAT scan or MRI of the brain
2 at
baseline. The follow-up had to be done
with the
3
same test a month after whole brain radiation day
4 10,
3 months after day 10 and every 3 months
5
thereafter until progression. All
CAT scans and
6
MRIs were centrally and independently reviewed by a
7
team of radiologists at the Neuroimaging Core
8
Laboratory at the Cleveland Clinic.
The reviewers
9
were blinded to study arm and treatment outcome.
10
[Slide]
11
Let me now review the results of RT-09, 5
12 38
patients were randomized in 82 sites in the
13
U.S., Europe, Israel, Australia and Canada; 267
14 patients
were randomized to the control arm and 271
15 to
the RSR13 arm.
16
[Slide]
17
The two arms were well balanced for
18
gender, RPA class, age and tumor type.
19
[Slide]
20
RSR13 did not impair the administration of
21
standard whole brain radiation therapy in this
22
population and 95 percent of the patients in the
235
1
control arm and 94 percent of the patients in the
2
RSR13 arm received all 10 doses of whole brain
3
radiation, with the mean number of doses in each
4 arm
of 9.9 and 9.8. Eighty percent of the
patients
5 in
the RSR13 arm received at least 7 doses of
6
RSR13, with a mean number of doses of 8.4.
7
[Slide]
8
According to the statistical analysis plan
9 and
following ICH guidelines, the primary
10
population for survival analysis was to be
11
comprised of the eligible patients.
Accordingly, a
12
blinded neuroradiology review was conducted to
13
determine eligibility and 22 patients were
14
identified in this review. In
addition, one
15
patient with small-cell lung cancer was also
16
excluded from this analysis.
Overall, this
17
represents a rate of ineligibility of only 4.3
18
percent.
19
[Slide]
20
The Kaplan-Meier curve shows the overall
21
survival for all eligible patients in this study.
22 In
yellow we see the RSR13-treated patients and in
236
1 red
the control arm. The median survival in
the
2
control was 4.4 months and in the RSR13 arm was 5.4
3
months. This represents a hazard
ratio of 0.7 by
4
unadjusted log-rank, and when these results were
5
updated with an additional follow-up of a year the
6
hazard ratio is consistent with the initial
7
analysis.
8
[Slide]
9
In the population of eligible lung
10
cancer/breast cancer patients, which is the
11
co-primary population for analysis, the median
12
survival in the control arm was 4.4 months with an
13
improvement of 38 percent, and a median survival of
14 6
months in the RSR13-treated patients. By
15
log-rank this is a hazard ratio of 0.81 with a p
16
value of 0.07. When these results
were updated
17
with an additional follow-up of a year the hazard
18
ratio is consistent with a p value of 0.05. In
19
yellow we see the RSR13-treated patients and in red
20 the
control arm, with an early separation of the
21
curves and separation through the median.
22
[Slide]
237
1
The protocol and the SAP specified the
2
conduction of a Cox multiple regression analysis
3
that had 17 prespecified covariates.
Of the 17
4
covariates, 7 were found to be predictive of
5
outcome in RT-09, and they are listed here. Those
6 7
covariates are KPS, extent of extracranial
7
disease, prior brain resection, primary site, age,
8
gender and baseline hemoglobin.
When all 17
9
covariates are incorporated in the model as
10
described in the SAP, the hazard ratio shows a 22
11
percent reduction in the risk of death in favor or
12 the
RSR13-treated patients, with a p value of 0.01.
13
[Slide]
14
In the eligible lung cancer/breast cancer
15
co-primary population the same analysis was
16
conducted following the SAP. The
covariates that
17
were predictive of outcome in this population were
18
KPS, extent of extracranial disease, prior
19
resection, age and gender. When
all 17 covariates
20 are
incorporated in the analysis the hazard ratio
21
shows a 24 percent reduction in the risk of death,
22
with a p value of 0.017.
238
1
[Slide]
2
In addition, to confirm the
results of the
3
Cox, we ran a stratified log-rank survival
4
analysis, including in this analysis the three
5
strongest covariates detected in the study. Those
6 are
KPS, prior resection and extent of extracranial
7
disease. When this analysis was
done in all
8
patients a hazard ratio of 0.81 is found including
9 all
three covariates, with a p value of 0.037.
In
10 the
non-small cell lung cancer/breast cancer
11
population the incorporation of just one covariate
12 in
the stratified log-rank shows a hazard ratio of
13
0.78, with a p value of 0.029.
14
[Slide]
15
Let me emphasize the results in the
16
eligible non-small cell lung cancer/breast cancer
17
co-primary population. In this
population we saw
18 by
unadjusted log-rank a hazard ratio of 0.81 with
19 the
corresponding p value of 0.07. The Cox
showed
20 a
24 percent reduction in the risk of death with a
21 p
value of 0.017. At this point the
logical thing
22 to
do was to look at the outcome of these two very
239
1
distinctive tumor types separately.
2
That is, indeed what we did. In
the
3
eligible non-small cell lung cancer patients the
4
log-rank showed a hazard ratio of
0.97 with the
5 Cox
showing a hazard ratio of 0.90. In
contrast, a
6
large treatment effect was observed in the eligible
7
breast cancer patients with a hazard ratio of 0.51
8 by
log-rank and a hazard ratio very consistent with
9
log-rank of 0.51 by Cox, both very consistent with
10
each other.
11
Let me emphasize that the eligible
12
patients with breast cancer do not represent an
13
arbitrary subset. They are the
result of a logical
14
analysis of the result that we encountered in a
15
co-primary population of lung cancer, breast cancer
16
patients.
17
[Slide]
18 This slide shows the overall
Kaplan-Meier
19
survival curve for the eligible breast cancer
20
patients. The median survival in
the control arm
21 was
4.5 months and in the RSR13 the survival was
22
doubled, to 9 months. By log-rank,
as we recently
240
1
reviewed, this shows a hazard ratio of 0.51 and by
2 Cox
the same hazard ratio with all 17 covariates
3
included in the analysis. In
yellow we see the
4
RSR13-treated patients and in red the control arm.
5
There is an early separation of the curves; clear
6
separation of the curves through the median and a
7
much larger number of long-term survivors in the
8
RSR13 arm.
9 [Slide]
10
In fact, we looked at the time of the
11
original analysis of the study for patients with a
12
survival of at least 12 months from randomization
13 and
this is what we encountered. Five
patients in
14 the
control arm had survived these 12 months.
Of
15
these, 3 had died at the time of the analysis. In
16
contrast, 11 patients in the RSR13 arm had survived
17 at
least 12 months from randomization and of these
18 9
were still alive at the time of the analysis.
I
19
would like to emphasize that all the survivors in
20 the
RSR13 arm had from adequate to excellent
21
performance status.
22
[Slide]
241
1
As I mentioned earlier, RT-09 was updated
2
with an additional follow-up of a year.
Therefore,
3 we
looked at all the breast cancer patients, in
4
this case with a minimum potential follow-up of 18
5
months by arm, and the results are shown here.
6
Each number represents an individual patient.
7
Those in white are patients that died at the time
8 of
the analysis; those in yellow are patients that
9 are
still alive. There were 7 patients in
the
10 control
arm that survived at least 18 months.
Two
11 of
these had died at the time of the analysis.
In
12
contrast, there were 15 patients in the RSR13 arm
13
that were alive a minimum of 18 months from
14
randomization. Of those, all those
in yellow were
15
still alive with survivals ranging from 18.5 months
16 to
almost 40 months, and there were 7 patients in
17
this column and 2 in this column with survivals in
18
excess of 2 years.
19
[Slide]
20
I will now focus on the
secondary
21
endpoints. Let me first point out
that by
22
statistical analysis planned the secondary
242
1
endpoints were to be analyzed in all randomized
2
patients.
3
[Slide]
4
Response rate in the brain defined per
5
protocol which is, in our view, the most important
6
secondary endpoint of the study considering that
7
RSR13 focuses on improving local therapy in the
8
brain, is shown here. There was
an 8 percent
9
difference in the response rate for all patients in
10
favor of RSR13. There was a 12
percent, and
11
statistically significant improvement in response
12
rate in the lung/breast co-primary population.
13
There was a 23 percent, statistically significant
14
improvement in response rate in the breast cancer
15
patients in the study. Let me
emphasize that all
16
those responses were determined by independent
17
radiologists.
18
[Slide]
19
RT-09 did not mandate confirmation of
20
response. Advice given at the
time the protocol
21 was
signed considered this impractical in a
22
population of brain metastases patients.
243
1
Therefore, we conducted an analysis that is not
2
planned in the protocol in patients that had a
3
follow-up CAT scan or MRI and minimum of 4 weeks
4
from the initial determination of response. We
5
defined that as confirmed response rate and the
6
results are shown here. There was
an 8 percent
7
difference in the rate of confirmed responses in
8
favor of the RSR13-treated arm.
There was a 9
9
percent advantage in the rate of confirmed
10
responses in the RSR13 arm in the lung/breast
11
co-primary population, and there was a 22 percent
12
difference in the confirmed response rate between
13 the
RSR13 and the control breast cancer patients.
14
[Slide]
15
In addition, we tried to explore the
16
impact of response and survival.
We looked at
17
responders and non-responders at 3 months and what
18
their subsequent survival was, and the results are
19
shown here. For patients that had
a PR or CR on
20 the
3-month scan, thus survival for those patients,
21 was
an additional 7.8 months. For
non-responders,
22
progressive disease and stable disease at the
244
1
3-month scan, those patients had an additional
2
median survival of 5.2 months.
3
[Slide]
4
We then compared the response rate at 3
5
months between the arms and those results are shown
6
here. In all patients there was a
7 percent
7
difference in favor of the RSR13-treated patients.
8 In
the lung/breast co-primary population there was
9 a
10 percent difference in favor of the
10 RSR13-treated
patients. In the breast cancer
11
patients there was a 13 percent difference in favor
12 of
the RSR13-treated patients.
13
[Slide]
14
Additional secondary endpoints for all
15
patients are shown here. There
was no difference
16 in
quality of life by KPS or Spitzer questionnaire,
17
cause of death, time to clinical or radiologic
18
progression between the two arms.
19
[Slide]
20
In the breast cancer patients there was a
21
significantly higher percentage of patients with
22
stable or improved KPS at 3 months or stable or
245
1
improved Spitzer questionnaire at 3 months in the
2
RSR13 arm. There was no
difference in cause of
3
death, time to clinical or radiologic progression
4
between the arms.
5
[Slide]
6
Clearly, we observed in this study a
7
different treatment effect of RSR13 in breast
8
cancer patients and lung cancer patients. This
9
difference could be due to many factors, some of
10
which are summarized here and they include
11
biological differences between these two very
12
different tumor types; different growth rates; and
13
differences in efficacy of they for extracranial
14
disease in these two tumor types.
One thing we
15
observed is that there are body weight differences
16 in
the distribution of high weight/low weight
17
patients between the arms and this may have
18
influenced the pharmacokinetics of RSR13,
19
specifically maximal concentration in the red
20
cells.
21
[Slide]
22
As we see here, when we classify patients
246
1
based on body weight, the RSR13-treated patients by
2
primary site and gender, we can see that the
3
majority of lung cancer patients are in the low
4
weight category independent of gender, and less
5
than half of the patients with breast cancer are in
6 the
low weight category.
7
[Slide]
8
We then studied the pharmacokinetics of
9 RBC
by body weight, tumor type and dose,
10
specifically RSR13 RBC concentration which is the
11 key
parameter because this is the site of action
12 of
RSR13, and we observed that patients in the lung
13
cancer low weight category that received 75 mg/kg
14 has
a lower median concentration in the red cell
15
than any of the other groups studied.
If you
16
remember from Dr. Kavanaugh's presentation, this
17
median concentration in the red cell will be below
18
what would be expected to generate the desired
19
pharmacodynamic effect through RSR13.
20
[Slide]
21
Let me summarize the efficacy data before
22 we
review the safety results. We saw
significant
247
1
reduction in the risk of death in the prespecified
2
co-primary populations by Cox multiple regression.
3 We
saw an improvement in response rate and a 38
4
percent improvement in the median survival time in
5 the
eligible lung cancer/breast cancer co-primary
6
population. In the eligible
breast cancer patients
7 we
saw an improvement in response rate; a
8
clinically meaningful improvement in survival with
9 a
doubling of the median survival; and a higher
10
number of long-term survivors in the RSR13 arm.
11
[Slide]
12
Let me review the safety profile of RSR13,
13
focusing on the result of RT-09.
14
[Slide]
15
First let me say that more than 500
16
patients to date have received RSR13 as an adjunct
17 to
radiation therapy in a series of Phase 1, 2 and
18 3
studies that are listed in this slide.
These
19
patients have received anywhere from 2-32 doses of
20
RSR13 and a dose of RSR13 has been up to 100 mg/kg.
21
[Slide]
22
One important point is the issue of
248
1
hypoxemia which is the most characteristic adverse
2
event related to the use of RSR13.
If you recall,
3 the
CTC grading scale defines supplemental oxygen
4 as
a grade 3 toxicity. By protocol, all
these
5
patients were on supplemental oxygen, therefore, we
6 had
to design a hypoxemia grading scale that was
7
adequate for these studies, and that is shown here.
8
This scale uses the length of oxygen
9
supplementation, the flow of oxygen required, and
10 the
presence or absence of symptoms or requirement
11 for
hospitalization to grade high hypoxemia.
It is
12
important to point out that grade 4 hypoxemia in
13
this grading scale is the use of CPAP or mechanism
14
ventilation and that is identical to the CTC scale.
15 Of
note, there were no grade 4 hypoxemic adverse
16
events in RT-09.
17
[Slide]
18
This slide shows treatment-emergent
19
adverse events that occurred in at least 20 percent
20 of
the patients in RT-09, all patients by arm and
21 the
breast cancer patients by arm. The ones
22
highlighted in yellow are those that were
249
1
significantly higher in the RSR13-treated patients
2 and
they include headache, nausea, hypoxemia,
3
vomiting and infusion symptoms.
However, the
4
majority of adverse events were grade 1 and 2.
5
[Slide]
6
This table lists the grade 3 adverse
7
events that occurred in more than 5 percent of the
8
patients, once again by arm and in the breast
9
cancer patients by arm. The most
frequent grade 3
10
adverse event in all patients receiving RSR13 was
11
hypoxemia, with 11 percent. Let
me emphasize that
12
hypoxemia does not mean hypoxia in this setting.
13
This is either low saturation, longer requirement
14 for
oxygen or need for more than 4 L of oxygen to
15
maintain saturation, or one of the other factors
16
defined in the scale. This is not
tissue hypoxia.
17 The
most common grade 3 adverse event in the breast
18
cancer patients were nausea and vomiting, at 8
19
percent each.
20
[Slide]
21
Grade 4 adverse events were even less
22
common. These are grade 4 AEs
that occurred in
250
1
more than 2 patients by arm and in the breast
2
cancer patients.
3
[Slide]
4
Further emphasizing the role of RSR13
5
adverse events, we reviewed the drug-related grade
6 4
adverse events in RT-09 by primary tumor type.
7
There were no grade 4 drug-related adverse events
8 in
the breast cancer patients treated in RT-09.
9
[Slide]
10
Regarding hypoxemia, only 11 percent of
11 the
patients treated in RT-09 had a grade 3
12
hypoxemia adverse event. Of
these, 73 percent were
13
asymptomatic. Hypoxemia was
self-limited and
14
easily managed with supplemental oxygen in all
15
patients.
16
[Slide]
17
To summarize the safety, we have data from
18 535
patients that indicate that RSR13 is safe in
19
cancer patients receiving radiation therapy. We
20 saw
a very low incidence of grade 3-4 adverse
21
events in a heavily pre-treated population of
22
cancer patients in RT-09. All
Adverse events in
251
1
RT-09 resolved within the 1-month follow-up period
2 and
were easily managed with supportive care.
3
Hypoxemia associated with RSR13 is self-limited;
4
requires only supplemental oxygen and is
5
asymptomatic in the majority of the patients.
6
[Slide]
7
At this point, I would like to turn to the
8
microphone over to Dr. Paul Bunn.
Dr. Bunn is
9
Professor and Director of the University of
10
Colorado Comprehensive Cancer Center and he will
11
provide some concluding remarks.
Dr. Bunn?
12 Conclusions
13
DR. BUNN: Thank you, Pablo. ODAC
14
members, FDA staff and guests, as a clinician who
15
sees many patients with brain metastases, I am
16
pleased to share my views on these studies and
17
their results.
18
[Slide]
19 Clearly, brain metastases are
associated
20
with disabling symptoms and short survival in these
21
patients. This is an unmet
need. Having enrolled
22 538
patients, this study represents the largest
252
1
randomized, controlled study of its kind.
2
[Slide]
3
As shown in this slide, the survival in
4 the
non-small cell lung cancer and breast cancer
5
prespecified co-primary population was superior in
6 the
RSR13-treated patients, with a median of 6
7
months in the treated group compared to 4.4 months
8 in
the control group. This survival
represents a
9 19
percent reduction in the hazard ratio of death
10 by
log-rank and 23 percent by Cox multiple
11
regression analysis, with corresponding p values of
12 p
equals 0.07 and 0.02 respectively.
13
I would note as a clinician that the
14
log-rank p value in the final analysis with 12
15
months of additional follow-up is 0.05.
In my
16
opinion, not the statistician's, this represents
17 the
most important data as it has the most events.
18 I
would also note that the magnitude of the hazard
19
rate reductions are comparable to those induced by
20
approved cancer therapies, including
21
cisplatin-based chemotherapy for non-small cell
22
lung cancer. Thus, I consider
this study to be
253
1
positive in this prespecified co-primary group of
2
patients.
3
When the data were analyzed in the
4
non-small cell lung cancer and breast cancer
5
populations separately it became evident that the
6
breast cancer patients had the greatest survival
7
benefit, with a median survival of 9 months in
8
RSR13-treated patients compared to 4.5 months in
9
control patients. Breast cancer
patients also
10
benefited the most in the secondary analyses, with
11
statistically significant increases in objective
12
response rate, performance status, Spitzer
13
questionnaire and fraction of patients alive at 12,
14 18
and 24 months. Obviously, breast cancer
alone
15
subset was not prespecified other than by
16
stratification but garnered the most benefit.
17
With a positive survival benefit for the
18
lung/breast cancer co-primary population, but most
19 of
the advantage in breast cancer patients, would
20 it
be best to approve RSR for both types of
21
patients or for breast cancer patients alone? This
22 is
why we have ODAC and this is your decision.
254
1
Personally, I would vote for approval of the
2
prespecified lung/breast cancer patient co-primary
3
population.
4
However, given the fact that the results
5 in
the prespecified population were largely driven
6 by
breast cancer patients, I would feel comfortable
7
voting for approval in breast cancer patients
8
alone. I say this because of the
huge efficacy
9
benefit in breast cancer patients produced by RSR13
10
combined with an acceptable safety profile in a
11
heavily pre-treated population.
12
At this time I will turn the podium to Dr.
13
Cagnoni for questions.
14
DR. PRZEPIORKA: We will hold the
15
questions until after the FDA presentation. Dr.
16
Ridenhour?
17 FDA Presentation
18 Clinical Review
19
DR. RIDENHOUR: Good afternoon.
20
[Slide]
21
My name is Kevin Ridenhour and I will
22
present to you the results of the clinical review
255
1 for
this NDA.
2
[Slide]
3
All of these individuals assisted with the
4
review process. The presenters
for the FDA are
5
highlight. Following my report on
the clinical
6
portion of this NDA, Dr. Sridhara will discuss the
7
statistical issues.
8
[Slide]
9
I will briefly cover the regulatory
10
background of RSR13 and describe the two trials
11
submitted to support this NDA. I
will then discuss
12 the
findings from study RT-008. The
remainder of
13 the
discussion will focus on the RT-009 study.
14
[Slide]
15
The applicant's proposed indication for
16
RSR13 is as adjunctive therapy to whole brain
17
radiation for the treatment of brain metastases
18
originating from breast cancer.
19
[Slide]
20
In June, 1995 the IND for RSR13 was first
21
submitted. In June, 2003 we
discussed with the
22
applicant our concerns regarding the lack of a
256
1
survival benefit in RT-009 and our concerns with
2
their subgroup analysis. In July,
2003 the
3
pharmacology data was submitted as the first
4
component of the NDA. In
December, 2003 the
5
clinical and statistical components were received
6
finalizing the NDA submission.
7
[Slide]
8
The two clinical trials submitted to
9
support this NDA are RT-009 and RT-008.
RT-009 was
10 a
randomized, open-label study of standard whole
11
brain radiation therapy and oxygen, with or without
12
RSR13, in patients with brain metastases. There
13
were 267 patients on the control arm and 271
14
patients on the RSR13 arm.
15
RT-008 was a single-arm study of RSR13
16
administered to patients receiving standard whole
17
brain radiation therapy with oxygen for brain
18
metastases. There were 69
patients in this study.
19
[Slide]
20
In RT-009 patients on the RSR13 arm
21
received 100 or 75 mg/kg through central
22
intravenous infusion over 30 minutes daily within
257
1 30
minutes of whole brain radiation therapy.
Whole
2
brain radiation therapy was given as 30 Gy in 10
3
fractions.
4
Patients on the control arm received whole
5
brain radiation therapy given as 30 Gy in 10
6
fractions and at least 4 L/minute of supplemental
7
oxygen was given to both arms 35 minutes prior to,
8
during and for at least 15 minutes after the
9
completion of whole brain radiation therapy.
10
[Slide]
11
The primary endpoint in RT-009 was
12
survival in the overall population as described in
13 the
original protocol and subsequent versions.
14
With the second protocol amendment the applicant
15
provided the description for an analysis to be done
16 in
the non-small cell lung/breast co-population.
17 Dr.
Sridhara will also discuss these analyses
18
further in her presentation.
Secondary endpoints
19
included time to radiographic and clinical
20 progression in the brain, response rate in the
21
brain, cause of death and quality of life.
22
[Slide]
258
1
The major eligibility criteria were a
2 Karnofsky
Performance Status greater than or equal
3 to
70, radiographic studies consistent with brain
4
metastases, resting and exercise SpO-2 greater than
5 90
percent on room air. Concurrent steroid
therapy
6 was
allowed, and the presence of a cytologically
7
confirmed primary malignancy.
Patients with small
8
cell carcinoma, germ cell tumors and lymphomas were
9
excluded. In addition, patients
with
10
leptomeningeal spread were also excluded.
11
[Slide]
12
This slide illustrates the even
13
distribution of tumor histology across both
14
treatment arms. Non-small cell
lung cancer was the
15
most predominant type, followed by breast and other
16
subgroup, mostly melanoma, colorectal and renal
17
cell carcinoma.
18
[Slide]
19
In the overall population the distribution
20 of
post-randomization systemic treatment types
21
appear even between both study arms.
22
[Slide]
259
1
But in the breast subgroup subsequent
2
exposure to radiation therapy, chemotherapy and
3
hormonal therapy appeared slightly more frequent on
4 the
RSR13 arm.
5
[Slide]
6
The number of brain lesions appeared to be
7
fairly well distributed in the overall population
8
between the control arm and the RSR13 arm.
9
[Slide]
10
However, within the breast cancer subgroup
11 a
higher proportion of patients with 3 or more
12
brain lesions was noted in the control arm. The
13
distribution of patients with only 1 brain lesion
14 was
greater on the RSR13 arm. This suggests
the
15 presence
of a greater tumor burden in breast cancer
16
patients on the control arm which may have
17
influenced outcome.
18
[Slide]
19
I will now summarize the efficacy results
20 for
RT-009. There was no survival advantage
21
demonstrated in the overall population or in the
22
non-small cell lung/breast co-population. These
260
1
were the two prespecified populations for analysis
2 defined
in the protocol. After analysis of their
3
data, the applicant is claiming a survival
4
advantage in a non-prespecified breast cancer
5
subgroup which we consider exploratory at this
6
time. Again, Dr. Sridhara will
also discuss this
7
further during her presentation.
8
[Slide]
9
As previously discussed, one of the
10
secondary endpoints was response rate in the brain.
11 In
response to a query from the FDA during the
12
review process, the applicant stated that
13
confirmation of response was not required for
14
RT-009. However, the applicant
provided estimates
15 of
confirmed responses and this was done by
16
comparing the response of the first scan taken
17
after the dose response to the best response. If
18 the
response was the same as best response, the
19
response was considered confirmed.
This is
20
demonstrated under the confirmed column on this
21
slide. Whether you look at total
versus confirmed
22
responses between treatment groups, there is a
261
1
trend in response rate that favors the RSR13 arm
2 but
it is not statistically significant. The
3
confidence intervals do overlap.
4
[Slide]
5
This slide illustrates distribution of
6
neurologic and non-neurologic causes of death.
7
These findings show that the majority of patients
8
with brain metastases died of non-neurologic
9
causes, causes that were not influenced by RSR13.
10 The
results are a large number of indistinguishable
11
causes of death.
12
[Slide]
13
As expected, most patients on both
14
treatment arms received steroids.
The distribution
15 of
steroid use was comparable between both
16
treatment arms.
17
[Slide]
18
In addition to the fact that most patients
19
that did not die of neurologic causes, we have the
20 following concerns regarding the relevance of
the
21
response assessment.
22
Given that there is no apparent advantage in
262
1
response rate in the brain with RSR13, whole brain
2
radiation and oxygen versus whole brain radiation
3 and
oxygen, there does not appear to be a
4
contribution of RSR13 to tumor response.
More than
5 90
percent of patients in both arms received
6
steroids, and response duration cannot be assessed
7
since confirmatory imaging studies were not
8
required. Also, the designation
of complete
9
response and partial response was given
10
irrespective of the appearance of a new brain
11
lesion.
12
[Slide]
13
As for the other secondary endpoints, the
14
applicant found no statistically significant
15
difference between the control arm and RSR13 arm in
16
time to radiographic tumor progression introduction
17 he
brain, time to clinical tumor progression in the
18
brain and quality of life.
19
[Slide]
20
RT-008 was a single-arm study with 69
21
patients given RSR13 and whole brain radiation
22
therapy with oxygen. This
included mostly patients
263
1
with lung cancer and breast cancer.
The median
2
survival was reported as 6.4 months but in a
3
single-arm study it is difficult to interpret time
4 to
event points such as survival. Response
rate in
5 the
brain was 29 percent. However, in a
setting
6
where patients received RSR13, oxygen and radiation
7 the
relevance of this response rate is difficult to
8
interpret.
9
[Slide]
10
Moving on to safety in RT-009, RSR13
11
exposure was similar between the overall population
12 and
non-small cell lung/breast co-population.
13
Radiation exposure was also similar between the
14
overall population and non-small cell lung/breast
15
co-population. The FDA was able
to reproduce the
16
applicant's analyses for RSR13 and radiation
17
exposure.
18
[Slide]
19
As for oxygen exposure, patients on the
20
RSR13 arm appeared to have received a longer
21
duration of oxygen therapy than patients on the
22
control arm. We should note again
that oxygen is
264
1
hypothesized to be a modifier of the biologic
2
effect of ionizing radiation and, as noted in the
3
slide for oxygen exposure, some of the extreme
4
values observed for the duration of oxygen
5
delivered beyond 24 hours could be related to
6
hypoxia exacerbated by RSR13, requiring prolonged
7
oxygen delivery.
8
[Slide]
9
The treatment-emergent adverse events
10
shown on this slide occurred with more frequency on
11 the
RSR13 arm. Of specific interest are
hypoxemia,
12 41
percent RSR versus 4 percent control;
13
hypotension, 13 percent RSR versus 1 percent
14
control; and vomiting, 38 percent RSR versus 17
15
percent control.
16
[Slide]
17
This slide shows the most common grade 3
18 and
4 adverse events. Again, hypoxemia was
more
19
common on the RSR13 arm. There
are also more cases
20 of
acute renal failure seen on the RSR13 arm.
21
[Slide]
22
In conclusion, there was no survival
265
1
advantage demonstrated for the RSR13 arm versus the
2
control arm in RT-009. There was
no advantage
3
demonstrated for RSR13 versus control in secondary
4
endpoints. The most common
adverse events included
5
hypoxemia, hypotension, nausea, vomiting and
6
headache. Severe adverse events
also included
7
acute renal failure.
8
The exploratory analysis demonstrating a
9
survival advantage in the breast cancer subgroup,
10
consisting of 60 patients on the RSR13 arm and 55
11
patients on the control arm, is being further
12
evaluated by the applicant in a randomized study.
13
[Slide]
14
Now Dr. Sridhara will discuss the
15
statistical issues of this NDA.
Thank you.
16 Statistical Review
17
DR. SRIDHARA: Thank you, Dr.
Ridenhour.
18
Good afternoon. I am Rajeshwari
Sridhara,
19
statistical reviewer of this application.
20
[Slide]
21
In this presentation I will be focusing
22
only on the efficacy results of the confirmatory
266
1
registration study, RT-009. There
are three major
2
areas of concern in this application.
They are
3
overall finding, subgroup findings and multiplicity
4
issues. I will present the
concerns in each of
5
these areas in the following slides.
6
[Slide]
7
First with respect to overall finding,
8
evidence of efficacy has not been established.
9
Multiple analyses have been conducted and there
10
appears to be a lack of internal consistency in the
11
results.
12 [Slide]
13
Regarding the evidence of efficacy as
14
presented by the applicant, the median survival was
15 4.5
months and 5.3 months respectively in the
16
control whole brain radiation arm and the treatment
17 arm
with RSR13 followed by radiation. Of
note, the
18
study RT-009 was designed with an estimated median
19
survival of 4.5 7 months in the control arm. The
20
study was adequately powered to detect a difference
21 of
1.6 months in median survival in the overall
22
study population. As presented
here, there was no
267
1
statistically significant difference between the
2 two
treatment arms.
3
[Slide]
4
The two sets of results presented in the
5
previous slide correspond to the first one which
6
refers to the data submitted at the time of
7
application to the agency, which had a data cutoff
8
date of January, 2003. Subsequently,
the applicant
9
submitted updated survival data in March of 2004
10
which included updates up to January, 2004. Also,
11 it
should be noted that the p values presented here
12 are
not adjusted for multiple looks of the data and
13
these p values, as such, should not be compared to
14
0.05.
15
[Slide]
16
The applicant has conducted numerous
17
adjusted analyses, adjusting for many covariates
18
using Cox regression models.
These adjusted
19
analyses can only be considered as supportive when
20 the
overall unadjusted finding is positive.
As
21
stated in the ICH-E9 guidelines, in most cases
22
subgroup analyses are exploratory and should be
268
1
clearly identified as such. They
should explore
2 the
uniformity of any treatment effects found
3
overall.
4
[Slide]
5
The applicant had clearly stated that the
6 primary
analysis would be based on unadjusted
7
log-rank test and, in fact, had identified both in
8 the
protocol and subsequent statistical analysis
9
plan that the adjusted analyses would be considered
10
only as exploratory. The quote
from the
11
applicant's statistical plan reads as follows,
12
"while designated prospectively, supporting
13
analyses should be considered exploratory in
14
nature, and inferences made based on p values
15
should be done so with caution.
Primary reasons
16 for
exploratory analyses are for estimation rather
17
than hypothesis testing."
18
[Slide]
19
The applicant had stated in the original
20
protocol and its amendments under the section
21
"survival" that, "RPA class of primary cancer and
22
other important covariates, such as primary tumor
269
1
control, age, presence of extracranial metastases,
2
baseline KPS and number of metastatic lesions will
3 be
included in a multivariate Cox model, along with
4 the
treatment to test the relative importance of
5
these factors for survival."
6
[Slide]
7
These covariates are listed as protocol
8
covariates in this table.
Subsequently, the
9
applicant included the 18 covariates listed in this
10
table under SAP covariates in their final
11
statistical analysis plan under the section of
12
covariates and significance, with a comment that
13
these are exploratory in nature and the primary
14
reason for such analyses were for estimation and
15 not
hypothesis testing.
16
[Slide]
17
Here I will present results of one such
18 exploratory model. In this exploratory model I
19
have included the protocol-specified exploratory
20
analysis with the evaluating covariates, RPA Class
21 I
versus II; site of primary breast and non-small
22
lung cancer; primary control, yes/no; age group
270
1
less than 65 versus greater than or equal to 65;
2
presence of extracranial metastases, yes versus no;
3 KPS
group more than 90 versus less than 90; and the
4
number of brain lesions, single versus multiple.
5
It should be recognized that in
6
determining the RPA class for a given patient KPS,
7
age, whether or not primary was controlled and
8
extracranial metastases were present or not were
9
considered and these factors are likely to be
10
correlated.
11
[Slide]
12
This table lists the results of analysis
13 of
data submitted at the time of the application
14 and
analysis of updated survival data.
Within each
15 of
these data time points two sets of data have
16
been analyzed. One data set
consists of all
17
patients as randomized and the second data set
18
consists of only eligible patients.
19
The applicant, in their
statistical plan
20
which was finalized after the completion of
21
enrollment, had stated that these adjusted analyses
22
would be conducted in eligible patients.
Hence,
271
1
analyses in both data sets are presented here.
2
None of the analyses presented here demonstrated a
3
statistically significant treatment effect, as seen
4 in
this table.
5
The applicant has conducted Cox regression
6
analyses including 17 of the 18 covariates that
7
were added in the final statistical analysis plan.
8 The
applicant has submitted 48 Cox regression
9
models with the same 17 covariates, but varying
10
some covariates between a continuous variable and a
11
dichotomous variable. For
example, two models are
12
considered, one with age as a continuous variable
13 and
another with age as two groups, less than 65
14
years versus more than 65 years.
None of these
15
models were adjusted for multiple analyses.
16
[Slide]
17
In summary regarding the overall finding,
18 the
single, randomized RT-009 study conducted in
19
patients with brain metastases does not demonstrate
20
substantial evidence of benefit with respect to
21
survival in the overall randomized study
22
population.
272
1
[Slide]
2
The second area of concern is subgroup
3
findings. I will be presenting
results from two
4
subgroups, namely, non-small lung cancer/breast
5
primary subgroup which was added on as a co-primary
6
hypothesis during the course of the study, and the
7
second subgroup of patients with breast cancer
8
primary, which was a post hoc data-dependent
9
exploratory subgroup analysis.
10
The reason given by the applicant to have
11 a
co-primary hypothesis in the subgroup of
12
non-small cell lung cancer/breast primary patients
13 was
that this subgroup was a large homogenous
14
subgroup. Also, with the addition
of this
15
co-primary, the protocol was amended so that the
16
type-1 error rate was adjusted using a modified
17
Bonferroni procedure in order to maintain an
18
overall type-1 error rate of 0.05.
19
[Slide]
20
The results of comparison of survival
21
distributions in the subgroup of lung/breast
22
primary patients are presented in this slide.
273
1
Again, two sets of analyses were conducted with the
2
data submitted at the time of the application with
3 the
updated data. In both analyses the
median
4
estimated survival was 4.5 months in the control
5 arm
and 5.9 months in the RSR13 arm. There
was no
6
statistically significant difference between the
7
control and the RSR13 in both analyses.
8
The applicant submitted earlier data on
9
eligible patients only. The
protocol specified
10
that the primary analysis in the overall
11
population, as well as in the lung/breast primary
12
subgroup would be conducted in all patients but the
13 Cox
analysis would be done in eligible patients.
14
[Slide]
15
In summary, the single, RT-009 study
16
conducted in patients with brain metastases does
17 not
demonstrate substantial evidence of benefit
18 with
respect to survival in the subgroup of
19
patients with lung or breast primary cancer. Once
20
gain, the p values listed here should not be
21
compared to 0.05.
22
[Slide]
274
1
The findings of the non-prespecified
2
subgroup with primary breast cancer has three major
3
problems, namely, absence of overall survival
4
benefit; a very small subgroup; and apparent
5
imbalances. I will go over each
of these issues.
6
[Slide]
7
In the absence of overall survival
8
benefit, any subgroup advantage is questionable.
9 The
ICH-E3 guidelines clearly state that these
10
analyses are not intended to salvage an otherwise
11
non-supportive study but may suggest hypotheses
12
worth examining in other studies.
13
[Slide]
14
The second issue of concern is that the
15
breast primary subgroup is a very small group with
16 a
total of 115 patients representing only 21
17
percent of the study population, with 55 patients
18 in
the control arm and 60 patients in the RSR13
19
arm. Of these patients, 6 in the
control arm and 2
20 in
the RSR13 arm were ineligible according to the
21
protocol entry criteria. There
was a total of 7
22
patients who were misclassified at randomized, 6
275
1
patients who died in less than 1 month after
2
randomization, and there were 6 patients in the
3
RSR13 arm who received up to 2 doses only of RSR13.
4
These patients continued further to receive
5
radiation as in the control arm.
6
[Slide]
7
Furthermore, some imbalances were observed
8
between the two treatment arms in some baseline
9
factors and post-therapy factors, as presented by
10 Dr.
Ridenhour. Of those imbalances in a few
11
important factors are presented here.
Although
12
none of these factors were individually
13
statistically significant, it is not plausible to
14
determine the collective influence of these
15
imbalances to the subgroup findings.
16
[Slide]
17
Although we considered this as an
18
exploratory analysis only, this slide presents the
19
breast subgroup finding. As
presented by the
20
applicant with data as of the NDA submission, the p
21
value in this small subgroup of breast primary
22 patients
was 0.006. However, with the updated
276
1
survival data submitted by the applicant in March
2 of
this year, the p value has diminished to 0.02.
3 Of
course, we do have a problem in interpreting
4
these p values.
5
[Slide]
6
In summary regarding the subgroup of
7
patients with primary breast cancer, some
8
imbalances were observed and a true finding cannot
9 be
isolated. There appears to be no
robustness in
10 the
subgroup finding. The p values presented
in
11 all
these analyses are not adjusted for
12
multiplicity and, at best, given the lack of an
13
overall finding, this subgroup finding is
14
exploratory and hypothesis generating.
15
[Slide]
16
The third major area of concern in this
17
application is multiplicity.
There are three types
18 of
multiplicity concerns. First, multiple
19
hypotheses were tested. The
type-1 error rate was
20
only allocated for two hypotheses, one in the
21
overall population and the other in the lung/breast
22
subgroup. However, several
hypotheses were tested.
277
1
Also, multiple analyses of the same hypothesis were
2
conducted at different times and different
3
analyses. Unadjusted and adjusted
analyses were
4
conducted. Furthermore, multiple
subgroups were
5
also examined. None of these
analyses were
6
adjusted for multiplicity.
7
[Slide]
8
In this slide I would like to present some
9
important points to be considered when evaluating
10
results from a single study. it
is known that
11
inherent variability may produce a positive trial
12 by
chance alone. That is, a p or 0.05
implies that
13
1/40 studies of ineffective drugs will be positive.
14
The FDA guidance to industry also states
15
that it is critical that the possibility of an
16
incorrect outcome be considered and that all the
17
available data be examined for their potential to
18
either support or undercut reliance on a single
19
multicenter trial. Statistical
persuasiveness can
20
only be verified by replication, especially when
21 the
results under consideration are from a small
22
subgroup of patients.
278
1
[Slide]
2
Finally, here is a review of
results
3
presented. The applicant has
submitted results
4
from a randomized, controlled, open-label
5
multicenter single trial. The
analyses of these
6
results do not demonstrate efficacy based on the
7 primary endpoint of overall survival both in
the
8
overall population and in the subgroup of non-small
9
cell lung or breast primary patients.
Also, no
10
significant benefit was observed in any of the
11
secondary efficacy endpoints.
12
[Slide]
13
The apparent survival benefit claimed by
14 the
applicant in a small subset group of breast
15
cancer primary patients is questionable because of
16
imbalances possibly influencing treatment effect,
17
very small sample size from a single study, and
18
results of a post hoc exploratory analysis. Thank
19
you.
20 Questions to the FDA and the
Sponsor
21
DR. PRZEPIORKA: We will have
questions to
22 the
FDA and the sponsor. Dr. George?
279
1
DR. GEORGE: I have a question for
the
2
sponsor. The trial that was
mentioned as ongoing,
3
randomized trial, did I miss something here? Did
4 you
address that at all or could somebody tell us
5
what that is about?
6
DR. CAGNONI: The question is
about the
7
ongoing randomized trial. It is a
randomized trial
8 in
patients with breast cancer and brain
9
metastases.
10
DR. GEORGE: Is it exactly like
this one?
11
DR. CAGNONI: It is a very similar
study,
12
yes. It is focused on patients
with breast cancer
13 and
is very similar. Patients are randomized
to
14
RSR13 and no RSR13. Both arms
receive supplemental
15
oxygen and the primary endpoint is survival.
16
DR. GEORGE: What is the target
sample
17
size in that?
18
DR. CAGNONI: It is 360 patients.
19
DR. GEORGE: And where is it in its
20
conduct right now?
21
DR. CAGNONI: Twenty sites have
been
22
initiated in the U.S. and Canada and patients are
280
1
being enrolled.
2
DR. PRZEPIORKA: Dr. Mortimer?
3
DR. MORTIMER: I am just curious,
in the
4 new
study have you stratified for estrogen receptor
5 and
HER2 status, and do we happen to know that in
6
this present study at all?
7
DR. CAGNONI: The ongoing study
stratifies
8 by
liver metastasis and KPS which were the two
9
strongest prognostic factors in RT-09, in addition
10 to
resection which is not allowed in the current
11
study.
12
DR. MORTIMER: Was HER2 known in
RT-09?
13
DR. CAGNONI: No, it was not.
14
DR. MORTIMER: So, you don't know
that it
15 is
not a prognostic factor.
16
DR. CAGNONI: There isn't a lot
of
17 literature
on the subject. The very little there
18 is
out there doesn't seem to indicate that there is
19 a
difference in survival in HER2-neu versus HER2
20
positive versus negative patients once they develop
21
brain metastases. What we do have
from RT-09 is
22 the
percentage of patients that received
281
1
trastuzumab after randomization, and those numbers
2 are
roughly similar between the arms.
3
DR. PRZEPIORKA: Ms. Portis?
4
MS. COMPAGNI-PORTIS: Yes,
considering
5
that that study is recruiting and accruing at this
6
time, why aren't we waiting for those results? Why
7 are
we looking at this now?
8
DR. CAGNONI: We fully believe that the
9
data that we have presented today is sufficient for
10
approval of RSR13 in patients with breast cancer
11 and
brain metastases. That study is in the
process
12 of
being initiated. It could take a very
long
13
period of time to accrue 360 breast cancer
14
patients.
15
MS. COMPAGNI-PORTIS: How long do
you
16
think that will be?
17
DR. CAGNONI: I can't
speculate. I can
18
tell you that it took 29 months to enroll 115
19
breast cancer patients in the study we are
20
reviewing today.
21
MS. COMPAGNI-PORTIS: Thank you.
22
DR. PRZEPIORKA: Dr. D'Agostino?
282
1
DR. D'AGOSTINO: I don't want to
stop the
2
discussion on the new study but I have a different
3
question. I am just a simple
statistician from
4
Boston so maybe I am off but it seems to me like
5 the
sponsor keeps claiming that they have
6
significant results, especially with the addition
7 of
data, and the FDA does not. Could we
have an
8
agreement? Is this significance
on the overall in
9 the
subset or is there not significance,
10 statistical
significance?
11
DR. PRZEPIORKA: If I can just
rephrase
12 the
question, is it true you have shown both in 008
13 and
009 that the median survival for the breast
14
cancer subgroup is doubled and significant?
15 DR. D'AGOSTINO: Well, even in the
16
overall--there is a slide on page 28 and the p
17
values are 0.05 for overall survival in breast, and
18 I
think somewhere there are also sheets that have
19
significance or other survival.
My understanding
20
from what the FDA is saying and reading is that
21
there is not statistical significance with the
22
overall survival. Is that agreed
upon?
283
1
DR. CAGNONI: If we could have the
slide
2 up,
it summarizes the analyses we conducted
3
following the SAP.
4
[Slide]
5
The SAP specified eligible patients, two
6
co-primary populations, and this shows the
7 lung/breast co-primary population median
survival.
8 The
original analysis is in white, 4.4 months for
9 the
controls, 6 months for the RSR13. The
hazard
10
ratio is 0.81, the p value is 0.07.
By the
11
prespecified Cox multiple regression that was
12
conducted as the SAP described, the p value is
13
0.02.
14
DR. D'AGOSTINO: But I thought the
15
prespecified analysis was an unadjusted log-rank
16
test.
17
DR. CAGNONI: The primary analysis
was
18
unadjusted log-rank, correct.
19
DR. D'AGOSTINO: So, it is not the
0.02.
20
DR. TEMPLE: But it says eligible
up
21
there. That is where the
difference comes I
22
believe.
284
1
DR. CAGNONI: That is the
difference.
2
DR. TEMPLE: It would be good if
everybody
3
addressed that.
4
DR. D'AGOSTINO: That is what I
was going
5 to
get to, are we dealing with different analyses
6 or
are we dealing with different groups of
7
individuals?
8
DR. TEMPLE: Different analyses,
at least
9 in
part.
10
DR. PRZEPIORKA: Go ahead, Dr.
Sridhara.
11
DR. SRIDHARA: The analysis that I
12
presented was in the ITT population.
Those were
13 the
p values that I was presenting both in the
14
overall population as well as in the non-small
15
cell/breast cancer population.
The results that
16 you
are seeing, both 0.07 and 0.05 in the non-small
17
cell lung/breast subgroup are based on eligible
18
patients only. Even so, we
wouldn't consider that
19 as
significant since we are not comparing with 0.05
20
since there are multiple hypotheses.
21
DR. D'AGOSTINO: So, in either
sets of
22
data it is not significant.
285
1
DR. SRIDHARA: Correct.
2
DR. TEMPLE: But the reasons are
multiple.
3 It
is important to tease them out. I think,
Raji,
4 you
are saying with two co-primary endpoints you
5
don't test at 0.05, you test at something smaller
6 but
nobody is quite willing to say at what, I
7
gather. So, that is one issue.
8
The other issue is the intent-to-treat,
9 the
all patients, or the eligible and that needs to
10 be
discussed too. Does everyone agree that
ITT was
11 the
prespecified endpoint? Because, if that
is so,
12
then that matters.
13
DR. D'AGOSTINO: I was assuming
somebody
14
else would pick it up but if nobody does, I would
15
like to.
16
DR. TEMPLE: No, everybody needs
to pick
17 it
up.
18
DR. D'AGOSTINO: I mean, it is
usual that
19 you
have an ITT sample as the sample that you are
20
analyzing as opposed to some definition of
21
eligible.
22
DR. PRZEPIORKA: Dr. Cagnoni?
286
1
DR. CAGNONI: Yes, if I may have
Dr. Scott
2
address the issue, please.
3
DR. SCOTT: Actually, this takes a
very
4
standard design that we, within the RTOG, have used
5 for
quite some time and most of the cooperative
6
groups as well. That is, with a
multicenter
7
clinical trial such as this, we are going to have
8
retrospective ineligibilities that are going to
9
occur. The design of this study,
as specified in
10 the
protocol, adjusted the sample size by 5 percent
11 to
account for the ineligibility that was expected
12 to
occur. So, the definition that we have
always
13
used is that eligible patients as randomized will
14 be
analyzed.
15 DR. D'AGOSTINO: Is that unusual for the
16
FDA, to get that type of description?
17
DR. TEMPLE: Well, as a general
matter an
18
after the fact exclusion raises potential problems.
19 You
know, if you know exactly how it is done and
20
whether it is all blind, and stuff, that is one
21
thing. But if you don't know
exactly how it is
22
done there is always a concern whether someone is
287
1 eligible
or not has something to do with the
2
outcome. So, I don't think that
is usual but other
3
people who know more about it can tell me. It
4
wouldn't be usual in other clinical disciplines; it
5
would be quite unusual.
6 DR. PAZDUR: I would also like to point
7 out
that if one takes a look at the ineligible
8
patients there are almost three times as many in
9 the
control arm as they are on the RSR arm.
I
10
don't know if these were prospectively suggested or
11
stipulated in the protocol about leptomeningeal
12
disease, no measurable brain lesions, dural disease
13 due
to bone, small-cell carcinoma--I know the small
14
cell carcinoma was at least one patient but are the
15 other
ones prospectively stipulated in the
16
protocol?
17
DR. CAGNONI: That is correct,
these are
18 all
exclusions based on the protocol. The
SAP
19
provided additional level of detail.
In following
20 ICH
guidelines, all these ineligibilities were
21
determined on pre-randomization factors.
The
22
specific eligibility criteria in the protocol that
288
1
would be used to define ineligibility were also
2
specified in the SAP and that was the analysis that
3 was
conducted. The reviews for ineligibility
were
4
conducted blindly by the same team of radiologists
5
that conducted the response assessment.
6
DR. PRZEPIORKA: Dr. Buckner?
7
DR. BUCKNER: Just a question for
the FDA
8
statistics group, if you analyze just the eligible
9
patients do you agree that even with the primary or
10 the
co-primary, in either set, there is a
11
statistically significant difference in survival?
12
DR. SRIDHARA: The p values that
the
13
sponsor presented, we agree with those p values
14
but, again as I said, in the non-small cell/breast
15
subgroup the p value of 0.07 and 0.05, with the
16
multiple hypotheses that we are testing, will not
17 be
considered as significant.
18
DR. PRZEPIORKA: Dr. Pazdur, do
you have
19
additional comments? No? Dr. Redman?
20
DR. REDMAN: Just for clarification
21
purposes, to the sponsor, confirmed responses are
22
defined how?
289
1
DR. CAGNONI: Yes, the protocol
did not
2
mandate confirmation of response.
3
DR. REDMAN: Right.
4
DR. CAGNONI: So, what we did was
in the
5
responders, we looked at those responders that had
6 a
CAT scan or MRI at a minimum of 4 weeks from the
7
response to termination. We
looked at those
8
patients and there was a certain number of patients
9
that did have CAT scans confirmed in those
10
responses. But I want to make it
clear that that
11 was
not an analysis per protocol.
12
DR. REDMAN: Then back to the FDA,
there
13 is
a statement on your slide 19, looking at the
14
exact same numbers that the sponsor provided for
15
confirmed responses--you state that there is no
16
apparent advantage in response rate but you don't
17
give a p value. Not that I am big
on p values but
18 the
sponsor gives a p value which is significant,
19
using the exact same numbers.
20
DR. SRIDHARA: The p value is
0.06.
21
DR. REDMAN: The sponsor has the
same
22
numbers and has a p value of 0.02--exact same
290
1
numbers on their slide on page 33.
2
DR. CAGNONI: If we can have the
slide up?
3
[Slide]
4
Are you talking about confirmed responses?
5
DR. REDMAN: Yes.
6
DR. CAGNONI: In all patients
these are
7 the
confirmed response rates for the two arms,
8
non-small cell/breast co-primary and breast cancer
9
patients.
10
DR. REDMAN: I was looking at all
11
patients.
12
DR. CAGNONI: All patients is the
top row.
13
DR. PRZEPIORKA: Any FDA response?
14
DR. SRIDHARA: I think there were
some
15
slight number differences there.
Let me get to
16
that.
17
DR. PRZEPIORKA: While she is
doing that,
18 Dr.
Martino, did you have a question?
19
DR. MARTINO: Two questions, both
to the
20
sponsor. I need to understand
more clearly what
21 the
causes of death were in the two populations of
22
breast cancer patients. Can
someone answer that
291
1 one
first? Did they die of systemic
disease? Did
2 they die of brain-related issues? And, was there a
3
difference between them?
4
DR. CAGNONI: Yes, the specific
cause of
5
death, results were collected. We
asked the
6
investigators to define cause of death as
7 neurologic,
non-neurologic or indistinguishable.
8 The
problem with evaluating cause of death in these
9
patients, this is very complicated in this
10
population. Can I have the slide
up, please?
11
[Slide]
12
Let me show the results. The
protocol
13
defined cause of death was neurologic,
14
non-neurologic, indistinguishable or alive. RTOG
15
combines indistinguishable and neurologic and those
16
results are shown here. Using
this classification,
17 for
the control patients there were 49 percent
18
neurologic versus 39 percent in RSR13; 51 and 62.
19
However, what I am showing you is not the analysis
20 by
protocol. The protocol included a
category of
21
indistinguishable that had a high number of
22
patients.
292
1
Let me also add that at the time of this
2
analysis 21 of the 60 patients in the RSR13 arm
3
were still alive, making the interpretation unclear
4 at
this point. I would also like to ask, if
I may,
5 Dr.
Friedman who has experience in treating
6
patients with brain metastases, for his opinion on
7
cause of death as an endpoint in this population
8 and
the ability to discriminate cause of death.
9
DR. FRIEDMAN: To be blunt, I
don't think
10 we
can do it. I think that is such a
challenging
11
proposition that in trying to discern why a patient
12
with brain metastasis died--from neurological
13
complications, from systemic disease in at least a
14
third to 40 percent we simply can't tell.
15
DR. MARTINO: But I think those of
us who
16
treat this disease, and there are those of us in
17
this room besides the present speaker, oftentimes
18 can
tell a brain-related death from a liver- or a
19
pulmonary-related death. It is
not such an
20
impossible task although, I will grant you, there
21 are
patients where it is not so obvious. But
you
22
have answered my question reasonably well enough
293
1
that I am happy with that.
2
I have one more, please. In these
3
patients, I am assuming that this was, in fact,
4
first therapy for their brain metastases but what
5 was
allowed subsequently, because I am sure many of
6
these relapsed and other things were done? Were
7
there restrictions imposed on that?
8
DR. CAGNONI: Regarding the first
part of
9 the
question, prior therapy for brain metastases
10 was
not allowed, with the exception of resection as
11
long as the patient had measurable disease after
12
that resection, in other words, they were partial.
13
Regarding subsequent therapy, I will ask
14 Dr.
Elias, who is Director of the Breast Cancer
15
Program at the University of Colorado, to comment
16 on
that since he conducted the review.
17
DR. ELIAS: Slide up, please.
18
[Slide]
19
Just also to discuss the previous question
20
briefly, sometimes patients may die of systemic
21
disease but if they have uncontrolled brain
22
metastasis you are much less likely to offer them
294
1
further therapy. That is one of
the reasons for
2 the
imbalance in the subsequent treatment for the
3 RSR
versus control groups.
4
In any case, this is subsequent treatment
5
and, as you see, there is comparable amount of
6
systemic or subsequent brain metastasis therapy.
7
Clearly, our options after primary treatment are
8
quite limited.
9
[Slide]
10
This analyzes the percent of
patients who
11
received different types of subsequent therapy.
12
Again, there is a slight predominance of more
13
chemotherapy being given, although this is not
14
statistically significant but this also may relate
15 to
the somewhat better Karnofsky performance status
16 of
those patients. Very few patients got
brain
17
surgery or stereotactic radiation.
18
[Slide]
19
This is the percent of patients who
20
received further therapy in terms of number.
21
[Slide]
22
This is the balance between the control
295
1 and
RSR13 group in terms of the specific agents
2 that
we have seen.
3
DR. PRZEPIORKA: Before we go back
to the
4
FDA, I just want outcome re-ask the question that
5 was
posed before. If I recall correctly, you
have
6 now
shown from 008 and 009, two studies, that the
7 median
survival is doubled with RSR?
8
DR. CAGNONI: That is correct, 008
did not
9
quite double the survival.
10
[Slide]
11
It was 5.4 versus 9.7 in 008 and 4.5
12
versus 7.0. This is Class II
patients. There were
13
very few Class I breast patients in 008.
This
14
compares the Class II patients.
15
DR. PRZEPIORKA: Dr. Sridhara, did
you
16
find the information you were looking for?
17
DR. SRIDHARA: I believe the applicant
18
presented that in all patients unconfirmed
19
responses were 37 versus 45 and we agree with that.
20 The
p value is 0.067. However, in the
confirmed
21
responses--I don't have the percentages but I can
22
tell you that in the control arm there were 43 of
296
1 the
267 who had responses, and 61 of the 271, and
2 the
p value that we got was 0.06 versus what the
3
applicant has given here which is 0.02.
4
DR. PRZEPIORKA: Dr. Redman, does
that
5
answer your question?
6
DR. REDMAN: Was that because you
couldn't
7
confirm some of the responses they confirmed?
8
DR. DAGHER: Another point that may
be
9
attributed to a slight difference in numbers, and
10 we
can discuss this, is that when we queried the
11
sponsor on this issue of confirmation they actually
12
gave us three sets of possibilities for patients
13 who
may be considered "confirmed."
The first was
14 if
some scan after the baseline and then a
15
subsequent scan--if you had the sequence of a CR
16 and
then a CR, they called that a confirmed CR.
If
17 you
had at some point a PR and subsequently another
18 PR
confirmed, that was a PR. But they also
had
19
this middle category where if you were PR and then
20
CR--or I think it was CR and then PR, that is
21
right, so if you had a CR on one scan and then the
22
scan you got right afterwards was a PR, they
297
1
considered that I think a complete response. I
2
don't know that we would agree with that
3
assessment. So, there may be a
slight difference
4 in
the interpretation of that middle group.
5
DR. PRZEPIORKA: Go ahead.
6
DR. CAGNONI: May I make a
comment?
7
[Slide]
8
This is the response rate in the brain per
9
protocol. All these analyses we
are discussing
10
were not per protocol. The
protocol specified that
11
they had to be done a certain way and it was done
12 the
same way in both arms, was reviewed
13
independently and is statistically significantly
14 higher
in the lung/breast co-primary population.
I
15
would like to emphasize that.
16
DR. DAGHER: Also, I would like to
17
emphasize that the main point we were trying to
18
make is that, yes, the issue of do you have a
19 difference
between the two arms is a significant
20
issue but also with this endpoint of response rate
21 in
the brain, what are the factors that would give
22 you
certainty or uncertainty regarding the
298
1
findings? So, the main points
were this issue of
2
confirmation, which is only one of several; the
3
fact that you had steroids on board with most
4
patients, which was appropriate but is certainly an
5 element that causes uncertainty when you are
6
looking at scans, edema, etc.
7
The other two that Kevin mentioned, one of
8
which was the fact that the protocol-specified
9
criteria did not require absence of any new lesions
10
when response, either CR or PR, was called. For
11
that last one that I mentioned, and the sponsor
12
will probably comment as well, in terms of this
13
issue of not requiring absence of any new
14
lesions--that was a small number of patients. But
15 we
are just showing that there are several points
16
here that make us uncertain about the contribution
17 of
RSR to the response in this particular trial and
18 in
the particular subgroup for which benefit is
19 claimed.
20
DR. PRZEPIORKA: Dr. D'Agostino?
21
DR. D'AGOSTINO: My point is
probably lost
22 now
but I just wanted to make sure that it was
299
1
understood that when you ask about the two studies
2 in
the breast cancer results, in fact, we have to
3
remember that the first study was a registry
4
comparison but, more important, the second study
5 was
not a planned group that was actually looked
6
at. So, the statistical
significance, be it double
7 in
terms of magnitude, could be questioned or
8
should be questioned, and also the sample size--we
9 are
dealing with only 20 percent of the original
10
sample so we are getting down to a smaller subset.
11
DR. PRZEPIORKA: And I think I
asked that
12
question because it was significant and it was
13
reproducible. So, even though it
is not
14
statistically valid it is certainly striking.
15 DR. D'AGOSTINO: I don't know the history
16 but
why didn't they focus the second study on that
17
group, given that they had something with the
18
registry? Did they go back later
on and find out
19 it
was in the registry as opposed to designing the
20
study? It doesn't look like they
designed the next
21
study with that result. If you
give me enough
22
time, I probably will find a subgroup that is
300
1
significant in both samples also.
2
DR. PRZEPIORKA: Dr. Buckner?
3
DR. BUCKNER: I have three
questions
4
related to the response endpoint.
First of all,
5
there was a statement that looks like 13 of the 115
6
patients were not assessable for response and that,
7 in
fact, the median survival of those with missing
8
data was 0.99 months in the RSR arm and 2.7 in the
9
control arm. Is that
correct? More than 10
10
percent of your cases had missing data?
11
DR. CAGNONI: Yes.
12
DR. BUCKNER: The second regards
scans.
13
Were patients required to have identical type of
14
scans for comparison? For
example, response
15
assessed, CT compared with CT scan?
16
DR. CAGNONI: Correct.
17
DR. BUCKNER: And that was
prespecified in
18 the
protocol, that they must have the same type of
19
scan for comparison?
20
DR. CAGNONI: That is correct.
21
DR. BUCKNER: The third question, were
22
patients required to be on a stable dose of
301
1
corticosteroids prior too the baseline scan for a
2
certain period of time before they were assessed
3 for
response?
4
DR. CAGNONI: No. However, number of
5
patients on steroids, mean median dose, dose
6
adjustments, increases, tapers and length of
7
steroids in days was comparable between the two
8
arms. There were no differences,
as the FDA I
9
think implied in one of their slides.
10
DR. BUCKNER: Were there in fact
patients
11
that were called responders that had a new lesion
12 on
a subsequent scan?
13
DR. CAGNONI: A small percentage of
14
responders had new lesions.
15
DR. BUCKNER: What percentage was
that?
16
DR. CAGNONI: Four percent and six
17
percent. Cam we have the slide
up, please?
18
[Slide]
19
Six percent of patients in the control had
20 new
brain lesions and four percent in the RSR13
21
arm.
22
DR. BUCKNER: And those were still
302
1
considered responders?
2
DR. CAGNONI: That is
correct. Dr. Dagher
3
explained that the way the protocol was written,
4
response could be determined as a PR or CR even in
5 the
presence of new lesions. When the study
was
6
designed the sponsor was advised that reseeding
7
could occur from extracranial systemic disease and,
8
therefore, to assess truly their response in the
9
brain new lesions should not be accounted for. The
10
percentage of new brain lesions was very small and
11
there were no new brain lesions in the breast
12
cancer patients that received RSR13.
13
DR. BUCKNER: Thank you.
14
DR. PRZEPIORKA: What percentage
of the
15
patients had hemoglobinopathies such as sickle cell
16
anemia?
17
DR. CAGNONI: We did not screen
for
18
hemoglobinopathies. Hemoglobin
electrophoresis was
19 not
done.
20
DR. PRZEPIORKA: And is there any
21
information from the clinical studies to suggest
22
that the abnormal hemoglobins might react
303
1
differently or confer additional toxicities?
2
DR. CAGNONI: I will have Dr.
Steffen,
3
head of pharmacology/toxicology, answer that
4
question.
5
DR. STEFFEN: In laboratory
studies using
6
human sickle cells, fetal cells and adult normal
7
hemoglobins, red blood cells RSR13 has no effect on
8
rheologic activity and the p50 effect is similar
9
across all hemoglobin types studied.
10
DR. PRZEPIORKA: Dr. Buckner?
11
DR. BUCKNER: I am sorry, on the
response
12
criteria one other question, if I may, what
13
proportion of your patients were followed by CT
14
scan and what portion by MRI?
15
DR. CAGNONI: The majority were
MRIs. I
16
can't give you the exact number.
We can try to get
17 it
for you in a few minutes but the majority were
18
MRIs.
19
DR. PRZEPIORKA: Other
questions? Dr.
20
Temple?
21
DR. TEMPLE: I am sorry to be
dense about
22
this, it is really a question for both groups, when
304
1 you
modified the study to give yourself co-primary
2
endpoints you must have identified a critical alpha
3 for
each of the endpoints. It wouldn't be
the
4
usual 0.05 ones; you had two of them.
So, what was
5
it? That is one question.
6
The second is, was your primary endpoint
7 for
the primary analysis the intent-to-treat
8
population or the eligible patients population? It
9
must be in the protocol or the statistical
10
analysis, it must be somewhere.
If Raji disagrees
11
with that, I want to hear what the disagreement is
12
because I have the same problem Ralph does. We are
13
sort of talking beside each other.
14
DR. CAGNONI: We will have Dr.
Scott
15
comment on that.
16
DR. SCOTT: Sure. The analysis was
17
specified as eligible patients as randomized--in
18 the
protocol.
19
DR. TEMPLE: In the protocol?
20
DR. SCOTT: In the protocol. Beyond that,
21 the
appropriate adjustment here that we used in the
22
protocol basically states that we will take the p
305
1
values, order them and then compare the highest p
2 value
to 0.05. If that is not significant,
then
3 the
next highest p value to 0.25, and so on until
4 you
get down to statistical significance. In
other
5
words, if we have 3 p values and they may be
6
ordered as 0.13, 0.08 and then 0.05 or 0.02 or
7
0.017, somewhere around there, then we would adjust
8 the
p value because the first one was not
9
significant, the second one was not significant and
10
then the third one would be adjusted at 0.05
11
divided by 3, which would be 0.0167.
Does that
12
help?
13
DR. TEMPLE: I think so but by
that
14
standard--you only had two co-primaries.
It sounds
15
like that procedure would not leave, say, 0.05 for
16 the
small cell plus breast as significant.
17
DR. SCOTT: Right.
18
DR. TEMPLE: Would that be true?
19
DR. SCOTT: That is correct. Right, as
20
long as the overall one was not significant it did
21 not
leave 0.05 and we didn't make the connection
22
that the unadjusted log-rank at 0.05 for the
306
1
updated data analysis--we did not say that that was
2
statistically significant.
3
DR. TEMPLE: Let me be sure I get
this,
4 for
the total population that is not significant.
5
That is clear, even in the new adjusted one.
6
DR. SCOTT: Right.
7
DR. TEMPLE: And when you make
whatever
8 the
right correction is for the second co-primary,
9 the
lung/breast, that wouldn't be either.
Right?
10
DR. D'AGOSTINO: That is what I
was asking
11
before and I thought I got the answer that neither
12
would be significant.
13
DR. SCOTT: Right, and then the
contention
14
that we had, which was that we needed to make an
15
adjustment for the heterogeneity by using an
16
adjusted p value, an adjusted test such as either a
17
stratified log-rank or Cox analysis.
So a Cox
18
analysis, as defined in the protocol, was performed
19 and
that reaches statistical significance.
20
DR. TEMPLE: Without dismissing
it, that
21
wasn't identified as the primary analysis. I mean,
22
sometimes you do things that aren't specified, I
307
1
understand, but it was not the primary analysis.
2
DR. SCOTT: It was specified in
the
3
protocol though as a confirmatory type of analysis.
4
DR. TEMPLE: As exploratory, but
if you
5
fail on the others you don't usually do
6
exploratory. Wouldn't that be
true?
7
DR. SCOTT: Not necessarily. I don't
8
agree with that and I will explain why.
That is,
9
when we design these studies and we design the
10
trial with the log-rank and also a Cox analysis
11
with the intent to use that analysis, we know
12
through simulation analyses and in the statistical
13
literature that you lose power if there is a
14
heterogeneity in the data set.
Thus, the only way
15
that you can retain that power as designed through
16 the
parameters of the study is to do a Cox analysis
17 or
stratified log-rank.
18
DR. TEMPLE: But nothing stops you
from
19
having specified that as the primary analysis in
20
case there was heterogeneity. I
mean, it is not
21
commonly done but you could do that.
22
DR. SCOTT: Right. We could have done
308
1
that. When I was part of the team
that designed
2
this study, back in the late '90s and early 2000,
3 we
didn't have the heterogeneity simulations
4
performed. So, at that time what
we did was the
5
unadjusted log-rank. So, I really
believe that the
6
statistical literature has helped us along that way
7 in
showing that aside from stratification the way
8 to
adjust for the heterogeneity is also in a
9
stratified log-rank.
10
DR. TEMPLE: I am not sure anybody
would
11
disagree with you but when it is done after the
12
fact the implications are somewhat different.
13
DR. SCOTT: But it was specified
that we
14 would do that. I mean, it is not like we looked at
15 it
and we saw, oh gee whiz, we missed and we are
16
going to go back and do something different. We
17
actually did what we specified in the protocol.
18
DR. TEMPLE: But you do wish it
had been
19 the
primary analysis now, of course.
20
DR. SCOTT: No, but it was part of
the
21
primary analysis.
22
DR. TEMPLE: Not exactly.
309
1
DR. PRZEPIORKA: Dr. D'Agostino?
2
DR. D'AGOSTINO: It wasn't the
primary
3
analysis. It says
"exploratory" and you did make
4
protocol amendments along the way that were
5
accepted. If the statistical literature
informed
6 you
that that would have been a better analysis or
7
analysis to tie into the primary you had plenty of
8
opportunity to do it before the data set was
9
locked. So, I am really not
following the
10
statement that the decision was made years ago. To
11 me,
it is not the primary analysis.
12 Open Public Hearing
13
DR. PRZEPIORKA: Any other
questions from
14 the
committee? Hearing none, we are going to
move
15 on
to the open public hearing. We have one
speaker
16 and
I need to inform the group that both the
17
believe in a transparent process for information
18
gathering and decision-making. To
ensure such
19
transparency at the open public hearing session of
20 the
advisory committee meeting, the FDA believes
21
that it is important to understand the context of
22 an
individual's presentation. For this
reason, the
310
1 FDA
encourages the open public hearing speaker, at
2 the
beginning of your written or oral statement, to
3
advise the committee of any financial relationship
4
that you may have with the sponsor, its product
5
and, if known, its direct competitors.
For
6
example, this financial information may include the
7
sponsor's payment for your travel, lodging or other
8
expenses in connection with your attendance at this
9
meeting. Likewise, the FDA
encourages you, at the
10
beginning of your statement, to advise the
11
committee if you do not have any such financial
12
relationships at all. If you
choose not to address
13 the
issue of financial relationships at the
14
beginning of your statement it will not preclude
15 you
from speaking. Our first speaker is
Peggy
16
Wesselski.
17
MS. WESSELSKI: Good
afternoon. My name
18 is
Peggy Wesselski and I am a cancer survivor.
I
19
have been happily married for 28 years to my
20
husband, Fred. We have three
wonderful daughters,
21 one
of which is with me today, my oldest daughter,
22
Amanda.
311
1
I was first diagnosed with stage 4
2
inflammatory breast cancer. At
that time, my
3
youngest daughter was in the first grade. I never
4
asked God why me but I did say Lord, my girls need
5
me. And, after much prayer I
realized that my
6
girls would be fine with their daddy and with God's
7
help. After all, He could be with
them 24/7. He
8
would be a better caregiver than I could be. I
9
surrendered my illness to the Lord for His will to
10 be
done, not mine. He has been blessing me
ever
11
since.
12
I have a lot of stories I could tell you
13 but
we are here to talk about RSR13. It was
14
January, 2002 when it was discovered that the
15
cancer had spread to my brain.
There were five
16
tumors, one of which had fluid around it. Dr.
17
Gabriel Hardabaji is my breast oncologist. He was
18 out
that day and I received the results--I am
19
sorry, he was out that day and I received the
20
results from the MRI from Dr. Therialt who gave me
21 the
news. I had already survived a lung met.
but
22
this sounded more serious to me.
Dr. Therialt said
312
1
that I would qualify for a study which he highly
2
recommended.
3 Arrangements were quickly made for
me to
4 see
Dr. Eric Chang. First the research nurse
came
5 up
and sat beside me. Her name was
Chris. She
6
told me all about the study and explained that
7
originally she was allowed only ten patients. She
8
already had those ten patients but she had just
9
found out that she could have another ten. Chris
10
smiled at me and she said, "you'll be mu number
11
eleven." That said to me
that God had gone before
12 me and made provisions so that I could take
part in
13
this study.
14
Chris went on and told me that all
15
patients in this study would have whole brain
16
radiation and receive oxygen but that some patients
17
would receive a 30-minute drip which was RSR13.
18 She
informed me that the computer would randomly
19
pick who would receive the drip.
At that moment I
20
thought if this is a good drug I know I am going to
21 get
it. I could already see God's hand on
it.
22
Everything happened so quickly that day
313
1
while I was being set up for the study, I lay still
2 on
the table having my helmet made for radiation.
3 It
sounded like a dozen people were in the next
4
room discussing my case. I heard
my name a few
5
times. I lay there thinking how
blessed I was.
6
They were scurrying around as if I were a
7
celebrity.
8
Later Chris came back and let me know that
9 I
would, indeed, be receiving RSR13. The
10
treatments went well. It didn't
seem to cause any
11
side effects that I can remember.
I did have a lot
12 of
fatigue which my doctor told me that I would
13 experience.
After treatment I remember being
14
warned that my first MRI, which would be one month
15
later, would probably not show improvement because
16
radiation works down the road.
But one month after
17 the
treatment with RSR13 and radiation my first MRI
18 did
show improvement. Each MRI showed more
19
improvement until there was only slight evidence
20
that something was there.
21
It has been almost two and a half years
22 now
and I am doing well. I am going about my
314
1
normal activities, doing anything and everything
2
with my family, enjoying my life to the fullest.
3 My
youngest daughter, who is a freshman in high
4
school now, keeps me on the run.
I am so thankful
5 for
M.D. Anderson, for Dr. Chang and for the
6
clinical trial that God allowed me to be a part of.
7 I
am thankful that I was number eleven and that I
8
did, indeed, receive RSR13.
9
Through my experience in fighting cancer
10 for
eight and a half years, I have made friends
11
with many other breast cancer patients.
It is my
12
hope that if they develop brain mets. they will be
13
guarantied this same opportunity to receive RSR13
14
that I had. I truly hope that you
will recommend
15 to
the FDA that they approve RSR13 to make it
16
available for all my friends and for other patients
17
with brain mets. as well. Thank
you.
18 DR. PRZEPIORKA: Thank you.
We appreciate
19
your comments. Lenny Matthews has
asked to speak.
20 Is
Lenny Matthews here? No? Okay, we will
21
continue on and the next presentation is by Dr.
22
Stephen George.
315
1 Subgroup Analysis in Clinical
Trials
2
DR. GEORGE: Well, I am doing
something a
3
little different. I am not
speaking directly to
4
this application but giving a little, brief primer
5 on
some generally accepted methodologic principles
6 in
clinical trials as they relate to subgroup
7
analyses. It is, of course,
relevant to this
8
discussion today but also to other discussion we
9 have
on this committee.
10
[Slide]
11
First, what do we mean by subgroup
12
analysis? I think it has been
clear that it is an
13
analysis of treatment effects within subgroups of
14
patients on a clinical trial. The
first question
15
that arises is why would you want to do this? If
16 you
designed it to do an overall test, why don't I
17
just do that and go home? Well,
the answer is we
18 all
have a suspicion that maybe there is something
19
going on that the treatment effects are not the
20
same in all patients on the study so it is a
21
natural kind of thing and humans want to search
22
around and find these kinds of things.
316
1
[Slide]
2
How often are these done? Well,
this
3
first paper I found said that approximately 50
4
percent of reports of randomized clinical trials
5
contain at least one subgroup analysis.
Actually,
6
Pocock has done a more recent analysis where the
7
answer is more like 70 percent. I
am actually
8
surprised it is that low. When I
read the
9
literature I thought it was 100 percent.
10
The second quote came for I.J. Good, back
11 in
the '80s, who said that deciding on analysis
12
after looking at the data is dangerous, useful and
13
often done.
14
[Slide]
15
Now, what are the basic problems with
16
subgroup analysis? Well, the
first one you have
17
already heard a lot about. I will
go into this a
18
little more and explain what this means but the
19
first is increased probability of type-1 error (the
20
null hypothesis) when there is really nothing going
21
on. If we look around, we have an
increased chance
22 of
spotting something and that would be erroneous
317
1 in
that setting.
2
The second is a problem sort of in the
3
other direction. It is decreased
power or what is
4
called an increased type-2 error in the individual
5
subgroups when, in fact, the alternative hypothesis
6 is
true, say, for example if the overall truth is,
7
unbeknownst to us, that there is an effect overall
8 and
it is the same in all subgroups and if we start
9
looking at subgroup and we are going to find a lot
10 of
them that aren't significant and maybe make the
11
wrong conclusion in the other direction.
12
The last is what we have seen already,
13
that all of these kinds of things create great
14
difficulty in interpretation.
15
[Slide]
16
What I would like to do first is point out
17
what are some general assumptions behind doing
18
clinical trials in the first place.
Well, the
19
hypotheses that we are testing usually address an
20
overall or what might be called an average
21
treatment effect in the study population.
22 The second point about that is that
there
318
1 is
no assumption in this of homogeneity of effect
2
across subgroups. We are not
assuming that the
3
treatment effect is the same in all subgroups just
4
because we are doing an overall test.
But what we
5 are
generally assuming to be the case is that the
6
direction of that effect, not necessarily the
7
magnitude but the direction of the treatment effect
8 is
the same in all the subgroups. That is,
we
9
would be very surprised, because of the way we
10
determine eligibility criteria and set up the trial
11 in
the first place, if we saw a result that showed
12
that treatment A worked in this subgroup and
13
treatment B worked in that subgroup.
More likely,
14 we
would see that if there is an overall effect
15
treatment A might work better in some groups than
16
others, but it is all sort of in the same general
17
direction.
18
[Slide]
19
The implications of these kinds of
20
assumptions that are behind most clinical trials
21 are
that the overall treatment comparisons are of
22
primary interest, and that is really what we did
319
1 the
trial for. We can use stratification or
2
regression techniques to adjust the overall
3
comparison for subgroups or covariates if we wish
4
but, again, those should be specified clearly in
5
advance. Subgroup analyses
themselves are
6
generally of secondary interest as hypothesis
7
generating techniques for future studies.
8
[Slide]
9
I think the key point about these subgroup
10
analyses is whether they were planned or not. So,
11 I
have mentioned something here, the pre-planned
12
analyses or hypothesis-driven kinds of
13
analyses--the subgroup hypotheses are specified in
14
advance and supposedly, because we have done that,
15 we
can control the error rates or the error rates
16 can
in principle be addressed but, as I will show
17 you
in just a second, that is not always so easy.
18 It
is a tricky business even when it is
19
pre-planned. By the way,
pre-planned does not mean
20 you
just said ahead of time that we were going to
21
look something. That is not the
same as actually
22
pre-planning the analysis.
320
1
The second type of subgroup analyses are
2
unplanned analyses or what would be exploratory
3
analyses. These are either
analyses suggested by
4 the
data or an exhaustive search for differential
5 treatment effects by subgroups. This is often
6
called by the pejorative term as data dredging,
7
although that is perfectly reasonable, again, if
8 you
realize that what you are doing is generating
9
hypotheses.
10
The problem with the unplanned analyses is
11
that you have inflated error rates and, in fact,
12 you
don't know what those error rates are because
13 you
really haven't specified what you were going to
14 do.
15
[Slide]
16 There are a couple of things in the ICH
17
guidelines that address subgroup analyses directly.
18
Here is one from the guideline E3, which is on
19
publication results, and it says it is essential to
20
consider the extent to which the analyses were
21
planned prior to the availability of data. This is
22
particularly important in the case of any subgroup
321
1
analyses because if such analyses are not
2 pre-planned they will ordinarily not provide
an
3
adequate basis for definitive conclusions.
4
[Slide]
5
In guideline E9, which is on statistical
6
considerations, says clearly that in most cases
7
subgroup or interaction analyses are exploratory
8 and
should be clearly identified as such.
These
9
analyses should be interpreted cautiously. Any
10
conclusion of treatment efficacy or lack thereof or
11
safety based solely on exploratory subgroup
12
analyses are unlikely to be accepted.
13
[Slide]
14
What about these error rates?
What are we
15
talking about here? If you looked
at k independent
16
subgroups and there is really no difference in the
17
treatments, the probability of finding at least one
18 is
represented by this formula, here. For
example,
19 if
you used the 0.05 level and looked at 10
20
different subgroups your chance of finding at least
21 one
is 0.4; it is not longer 0.05.
22
[Slide]
322
1
Here is just a graph of that, showing that
2
this increases quite rapidly as a function of the
3
number of subgroups. This is when
you know the
4
number of subgroups.
5
[Slide]
6
So, what can we do about it?
Well, of
7
course, one way is to control error rates. Well,
8 for
planned subgroup analyses you can control the
9
overall type-1 error rate. One
conservative way is
10 to
use this thing that is often called a Bonferroni
11
correction, which is to simply divide the overall
12
error rate by the number of analyses you are going
13 to
do. Of course, that gives you a much
smaller
14
alpha level on each particular test.
15
In this case, the power or the probability
16 of
detecting real differences when they are present
17 is
sharply reduced in individual subgroups.
Of
18
course, for unplanned analyses we don't know k and
19 the
error rates are really unknown, as I have
20
already mentioned.
21
[Slide]
22
Here is a hypothetical example and I will
323
1
show you a real example of where this happened and
2 I
think caused some problems. Let's
suppose we
3
have two groups, experimental and control. Outcome
4 is
overall survival. The null median is 12
months,
5 meaning if there is really no difference in
these
6
treatments and all we are doing when we are
7
randomly assigning them is sort of randomly
8
assigning people to the same thing, we would expect
9
about 12 months.
10
Alternatively, if the experimental
11
treatment is working, let's suppose the median
12
would be 16 months long. That is
a 25 percent
13
reduction, 0.75 hazard ratio.
Let's suppose we do
14
this trial with 36 months accrual, 12-month
15 follow-up,
500 patients on this study. We want a
16
0.05 overall alpha level and suppose the power is
17
0.8. Now, we have a couple of
subgroups here.
18
There are males and females.
Let's suppose that 70
19
percent of them are males in this study, about 350
20
males and 150 females.
21
[Slide]
22
What could we do? Well, you could
do
324
1
subgroup tests with no adjustment--not a good idea
2 but
we could do it, and we use 0.05 in each of the
3 two
subgroups. The overall type-1 error rate
has,
4 of
course, jumped up. It is no longer 0.05;
it is
5
closer to 0.1. But also the
power, the ability to
6
pick up the difference in the males is only 0.64
7 and
in females it is only 0.33. In fact, the
8
probability that the correct conclusion is reached
9 in
both subgroups, males and females, if in fact it
10 is
true that there is this difference in both
11
subgroups is only about 20 percent, 0.21.
12
[Slide]
13
Let's say, okay, that is not too god but
14 at
least we want to control the type-1 error rate
15 so
we could do this sort of conservative thing I
16 suggested
before and divide by 2. So, we use 0.25
17 in
each subgroup and, therefore, the overall type-1
18
error rate is controlled. It is
less than 0.05.
19 But
now, because we have made it harder to reject
20 the
hypothesis in the subgroups, the power is about
21
half in the males and only about a quarter in the
22
females and the probability that the correct
325
1
conclusion will be reached when, in fact, there is
2
something going on is very poor.
So that is not
3
good. By the way, the only way to
fix this is to
4
have a very large sample size.
5
[Slide]
6
Now let me give you a real example where I
7
think this occurred in almost exactly that kind of
8
scenario. This is what I call the
aspirin example.
9 I
am not going to go into great detail here but in
10
1978 there was a publication by the Canadian
11
Cooperative Study Group of an excellently done and
12
well run clinical trial of aspirin and another
13
drug. I am just going to focus on
the aspirin.
14
This was published in 1978 in the New England
15
Journal of Medicine. Their
conclusion in the
16
abstract, and emphasized in the discussion, was
17
among men--among men, remember--men and women were
18 on
this study, the risk reduction for stroke or
19
death was 48 percent, whereas no significant trend
20 was
observed among women. We conclude that
aspirin
21 is
an efficacious drug for men with threatened
22
stroke.
326
1
[Slide]
2
Here is what this was based on.
The first
3 row
here gives males and the columns give aspirin
4 and
no aspirin. Among the males there were
85
5
events, strokes or deaths, 29 on the aspiring group
6 and
56 on the no aspirin group out of the total
7
number of subjects of around 406.
So, it is about
8 70 percent and a great predominance of
events were
9 in
the no aspirin group, indicating an advantage
10 for
aspirin. In females, in fact, the
advantage
11
seemed to go in the other direction.
If anything,
12
there were more strokes or deaths in the aspirin
13
group among females, only 29 events total and the
14
total number of subjects was only 179.
The total
15
number of events, if you just look at that, which
16 is
what the trial was designed to do, still favors
17 the
aspirin group.
18
[Slide]
19
If you translate that into things that we
20
like to look at on these trials, which is the risk
21
reduction in stroke or death, if you just look at
22
that first row again for males, the risk reduction
327
1 was
about 48 percent. That first column, by
the
2
way, is observed over the expected number of events
3 in
the categories. But the risk reduction was about
4 48
percent. That is a very dramatic risk
for
5
males, chi square value 8.2, p value 0.004, nominal
6 p
value. For females it actually increased
by 42
7
percent, a chi square, but not a significant
8
result. Overall the risk
reduction was about 30
9
percent and a barely significant result by the
10
usual criteria.
11
[Slide]
12
Now, ten years later a large meta-analysis
13 of
all results of various types of antiplatelet
14
treatments was published in which they concluded,
15
among other things, that overall allocation to
16
antiplatelet treatment reduced vascular mortality
17 by
15 percent and non-fatal vascular events, stroke
18 or
myocardial infarction, by 30 percent. I
don't
19
have time to go into the details but basically they
20
found there is no difference in males and females.
21
Aspirin worked, and it worked to reduce the
22
mortality approximately by what the Canadians got
328
1 in
their first study ten years earlier.
During
2
those ten years, what was the advice given to women
3 in
this situation? So, it can happen. There can
4 be
some real mistakes made in looking at subgroup
5
analyses.
6
[Slide]
7
What can we do about this? How do
we
8
interpret subgroup analyses? We
know they are
9
going to be done. Here are some
guidelines that
10
were presented several years ago--or some of them,
11 and
I didn't put all of them on here--to look for
12
when you are reading about subgroup analyses that
13 are
done. First, were there a priori
hypotheses
14
stated? As I mentioned, I think
that is the most
15
important one. Second, what is
the clinical
16
importance of the difference if it is really real?
17
Third, did they assess the statistical significance
18
properly? In some cases, if it
wasn't planned, of
19
course, this may be almost impossible.
Is there
20
consistency across studies? This
is important but
21 it
implies there is more than one study.
And, is
22
there any indirect supporting evidence either from
329
1
preclinical studies of other
theoretical reasons
2 why
you expect that subgroup to be different?
That
3 one
is probably a weak one. Humans are
remarkably
4
adapt at coming up with reasons for anything they
5
find.
6
[Slide]
7
One thing I wanted to mention briefly is
8 the
idea of a treatment-covariate interaction
9
because nobody has talked about that today. This
10 is
sort of a generalization of subgroup concepts.
11
Basically, the idea is you don't have to be really
12
talking about subgroups, identified groups of
13
people. You can use so-called
covariates that are
14
continuous. For example, if you
have age you don't
15
have to say age above 65/below 65 you can use it as
16
just a continuous variable. Then
you can use this
17 for
testing for what are known as
18
treatment-covariate interactions.
Basically, it
19
means does the treatment differ in the sense of
20
having an interaction with this covariate. There
21 are
quantitative interactions, which is what is the
22
most common kind of thing, where the treatment
330
1
effects are in the same direction but of different
2
magnitude, and qualitative interactions where the
3
treatment effects are actually in opposite
4
directions, which would be rare.
5
[Slide]
6
This simply indicates the kind of thing
7
that I am talking about. If you
have a control
8
treatment and a covariate, males and females again,
9 and
an outcome depending on which treatment group
10 you
are in, whether you are male or female, and an
11
interaction term, this beta-3, XZ.
So, if you look
12
across the rows here, female and male, the
13
treatment effect in females is beta-1; the
14
treatment effect in males is beta-1 plus beta-3.
15 So
the statistical test becomes one of simply
16
testing for beta-3. The reason I
am pointing this
17 out
at all is whether beta-3 is zero. If it
is not
18
zero then there is something going on.
19
[Slide]
20
So, what are some strategies we could use
21
when we are interested in subgroup analyses? First
22 of
all, we could design for the overall hypotheses
331
1 but
test within predefined subgroups. As I
have
2
already noted, that has a high overall error rates,
3 low
power in the subgroups and biased estimates.
I
4
haven't emphasized biased estimates but what
5
happens in these subgroups when you find a
6
difference is that it is known to be biased. That
7 is,
it is going to be larger on average than what
8 the
truth is because you searched and haven't found
9
it. This is not a good
thing. In other words, in
10 the
aspirin example you could have guessed that
11
that effect in the males was too high.
It was just
12
sort of implausible, and that is what happens when
13 you
look in these subgroups.
14
Second, we could design for the overall
15
hypotheses but test for prespecified
16
treatment-covariate interactions, which is what I
17
just mentioned in the last slide.
That I think is
18 a
good strategy but it has low power to detect even
19
modest interactions. The only way
around this is
20 to
get much larger studies, which is a depressing
21
point. So, there is nothing easy
there.
22
[Slide]
332
1
Third, we could design for the overall
2
hypotheses as before and conduct unplanned,
3 exploratory
analyses of subgroup differences.
4
This, of course, gives us unknown error rates.
5
That is why we really say this is a
6
hypothesis-generating exercise for future study.
7 It
doesn't mean it is wrong to do this.
There
8
isn't anything wrong with it, it is just that you
9
have to recognize it for what it is.
10
Last, we could actually design for
11
prespecified subgroups or interactions.
That
12
allows us to control for the error rates but
13
produces depressingly large studies that are often
14
almost impossible to do.
15
[Slide]
16
So, what is the conclusion from all this?
17 One
is that I think pre-planning is key. It
is
18
very important to think very clearly about what you
19 are
doing and how you are going to do it,
20
particularly in a regulatory setting.
You can get
21
away with this more if you are just trying to
22
publish a scientific paper, as people obviously do,
333
1 but
it is a lot more difficult in a regulatory
2
setting.
3
Second, we do need larger studies if we
4 are
really going to do proper subgroup analyses.
5 There
is actually no way around that, I don't
6
think.
7
Third, exploratory analyses are good for
8
hypothesis generating but really are not convincing
9 by
themselves. The last point is more than
one
10
study is very important for validation.
It would
11
make results much more believable if you find two
12
studies with a strong subgroup interaction. That
13 is
it.
14 Committee Discussion
15
DR. PRZEPIORKA: Thank you, Dr. George. I
16 do
wish you had given your presentation earlier.
17 It
would have assisted in our discussion but if we
18
have any questions for him, now would be the time
19 to
do so. Hearing none, we will take a
10-minute
20
break. If we can return here at
3:25, we can get
21
started with the questions.
22
[Brief recess]
334
1
DR. PRZEPIORKA: We are going to
get
2
started. We are now into the
question portion.
3
Thank you for the brief and unbiased questions for
4 the
afternoon. The committee has received a
copy
5 of
the questions and the data that is felt to be
6
germane.
7
When the primary analysis in the overall
8
study population is negative, subgroup analyses are
9
considered to be exploratory, i.e., not capable of
10
providing a conclusive finding.
Although there
11
could be exceptional cases, these analyses still
12 pose
multiplicity and potential bias problems.
13
So, question number one is, in fact, the
14
survival analysis in the overall population of the
15
randomized trial is negative. Do
the observed
16
survival results from the single study in the
17
subgroup of patients with metastatic to the brain
18
represent substantial evidence of RSR13 efficacy in
19
this subgroup?
20
We will first open the question up to
21
comments and at the end of the comments call the
22
vote. Any comments from the
committee? Dr.
335
1
Martino?
2
DR. MARTINO: Well, first of all,
I want
3 to
thank the sponsors for realizing that this is a
4
fairly serious set of circumstances that they are
5
dealing with and, you know, for all of those of us
6 who
take care of breast cancer patients as well as
7 all
the other people with brain metastases, that
8
someone is directing attention at this is laudable,
9 and
for that I am grateful to them.
10
This data is very meaningful to me because
11 it
is an area that I deal with a great deal so I
12
appreciate its importance, and I do have the sense
13
that there probably is something going on here
14
which is of value. The issue for
me is, is it of
15
sufficient value for us to change the way that we
16
practice oncology?
17
Because if an agent is approved several
18 things
follow that. One of the things is that
the
19
agent is then used for the population for which an
20
application is sought and given.
But more than
21
that occurs, and that is that clinicians who have
22
other patients for whom they mean to do the very
336
1
best start to then ask the question, well, if it
2
works in population A, surely it must work in B, C,
3 D
etc. So, then a generalization of a
behavior
4
occurs.
5
So, for all of those things to be allowed
6 one
has to assume a great deal of responsibility
7 and
thinking through not only the simple decision
8 of
this drug in this population but the
9
consequences that follow. I think
I simply want to
10
remind all of you that that is, in fact, what we do
11
when we make these decisions. It
isn't simply that
12 we
approve something for a patient population.
13
Medical behavior expands beyond that and we have to
14
take all of that into consideration here.
15
The other issue that is of great concern
16 to
me is that I realize this company has another
17
study that they have started in the population of
18
interest. If we decide today to
proceed with this,
19
what will happen to that trial?
Well, you all know
20 the
answer to that. You have seen it over
and over
21 and
over. The answer is that that trial will
not
22
accrue. We will never know an
answer which is
337
1
based on more substance than what we see today, and
2 so
that is the other responsibility that we have to
3
take on our shoulders.
4
DR. PRZEPIORKA: Dr. D'Agostino?
5
DR. D'AGOSTINO: I was embarrassed
to
6
raise my hand and said let somebody else raise the
7
first issue, but she stole my thunder.
This is an
8
unspecified subgroup. I realize
that you look back
9 at
the registry and see results but it is based on
10 18
breast cancer patients. We have this
study with
11 the
subgroup showing some real interest,
12
unfortunately not specified. Then
we have an
13
ongoing study which will be doomed if we make a
14
mistake by over-interpreting the results that we
15
have before us, and I think it really is an
16
over-interpretation even if there wasn't that other
17
study out there, and I am very excited that there
18
is. Reading too much into this
data I think is a
19
real problem. I think this really
is unspecified
20 and
is very problematic in how to interpret it.
21
DR. PRZEPIORKA: Ms. Portis?
22
MS. COMPAGNI-PORTIS: Yes, I would
just
338
1
like to say as a person living with breast cancer
2 and
also as a patient representative and someone
3 who
has an opportunity to work a lot with people
4
with metastatic disease that I know that even small
5
results can be significant to a patient or a few
6
patients and that that is important.
Yet, I think
7
that these results are too preliminary and I really
8
think it is important that this other trial goes
9
forward. I know that recruitment
for the trial has
10
already slowed down because this was brought before
11 the
FDA, and I think it is really important that
12
that study goes forward. So, I
think we always
13
need to let the science lead and I don't think we
14
have the data yet that we need.
Thank you.
15
DR. PRZEPIORKA: Dr. Buckner?
16
DR. BUCKNER: Looking at the data
we have
17 and
one of the problems that we have with subsets
18
plus the statistical issues is are we really
19
comparing apples with apples?
And, looking for
20
sources of real imbalance between the arms has been
21
alluded to generally but not quite specifically,
22 not
in a summary fashion. So, when I was
looking
339
1 at
this I basically went through what are the
2
factors that I thought favored the RSR arm and
3
balance in favor of RSR with nothing to do with
4
treatment efficacy; what favored the control; what
5
seemed to be balanced and what were the unknown
6
factors. All of these have been
alluded to but
7
just to list them briefly, there were several that
8
actually favored RSR13, specifically fewer brain
9
metastases in each patient and also less of the
10
bidimensional products, so basically less disease
11 in
the brain; less disease in extracranial sites
12 and
normal number of metastatic sites; more
13
systemic therapy really, more chemotherapy and more
14
hormonal therapy in the patients on the RSR13 arm.
15 Is
that because they had better outcomes going into
16 the
radiation treatment or better outcomes coming
17
out? That is hard to sort out. In fact, a
18
slightly better performance score in the RSR.
19
There was at least one meaningful variable
20
that I think favored the control, which is that a
21
better baseline mental status generally portends a
22
better outcome in patients with brain metastases.
340
1
Then there were a number balanced, as we know, RPA
2
class, post-RSR treatment of brain metastases, age,
3
distal metastases and, as Joanne pointed out,
4
several important unknowns--the ER and PR status,
5 the
HER2 status, the prior number and types of
6
chemotherapy.
7
But putting it all together, even if there
8
weren't the statistical issues of subgroup
9
analyses, it seems that there are some fairly
10
substantial imbalances that one a priori might
11
expect that the patients receiving RSR13 would have
12 a
better outcome regardless of whether the
13
treatment were effective or not.
14
DR. PRZEPIORKA: Dr. Redman?
15
DR. REDMAN: Just for my
clarification
16
because, no offense, Dr. George, I thought I
17
understood this and now I am not so sure. The
18
study pre-identified a group of breast cancer
19
patients and it was a stratification factor. Is
20
that correct? Or, was that done
after the trial
21 was
started? Breast and lung.
22
DR. CAGNONI: It was in the
original
341
1
protocol, stratification criteria, that is correct.
2
Breast cancer was a stratification criteria.
3
DR. REDMAN: The prespecified
subgroup was
4 the
combination of breast and lung as a co-primary
5
endpoint. So, you know, that
carries considerably
6
more weight than something you look at afterward.
7
DR. REDMAN: Right, but the study
was not
8
powered to see a difference between them.
9
DR. TEMPLE: Well, you can
stratify a lot
10 of
things--
11
DR. REDMAN: Right.
12
DR. TEMPLE: You may or may not
choose to
13
analyze your strata as a separate group.
That is a
14
decision you make in plotting out your analysis
15
plan. Of course, the groups get
smaller and
16
smaller, as Dr. George said, so at some point you
17
don't expect to win because, you know, if your
18
group is only--what?--one-sixth of the total you
19
would have to have a really huge effect to win so
20 you
don't usually expect to. But you may
want to
21 be
sure they are equally distributed in the two
22
groups so you could stratify and not analyze. But
342
1
then you might put it in a covariate analysis if
2 you
claim the covariate analysis as your primary
3
analysis, which you have heard some debate about.
4
DR. PRZEPIORKA: Dr. George, do
you have
5
comments?
6
DR. GEORGE: Yes, just a couple of
7
comments on that point. The
purpose of the
8
stratification is to get slightly more homogeneous
9
groups on the theory that in those groups they will
10
have sort of responses about the same, but still
11 you
are sort of doing an overall test as the
12
primary thing unless you have specified something
13
else ahead of time, which in this case was the
14
combination of two of those groups, I guess.
15
Anyway, you do that presumably to get a little more
16
precision in your result.
17
With respect to the other issue of
18
imbalances among groups, presumably part of this
19 was
addressed with the sponsor's analysis of doing
20
covariate adjustments of various kinds.
The issue
21
though for us has to do with that prespecification
22 of
whether it was primary or not because that also
343
1
becomes after a while fairly exploratory if it
2
wasn't pretty well laid out ahead of time.
3
DR. PRZEPIORKA: Dr. Cheson?
4
DR. CHESON: We are in a bit of a
5
conundrum here. Whereas I
completely agree with
6 Dr.
Martino's analysis that if we do approve this
7
drug that trial is dead, if we don't then it also
8
sends another message that perhaps, you know, we
9
were not in favor of this drug and the trial may be
10
dead as a result of that decision.
11
So, if the latter is the
decision of this
12
committee, then I strongly recommend that the
13
wording be exquisitely careful to encourage
14
participation and not to suggest that it was
15
because we didn't think there was something there
16 but
that it required additional support for the
17
approval.
18
DR. PRZEPIORKA: I am going to
take the
19
chair's prerogative and perhaps put some words into
20 Dr.
Temple's mouth. I remember the days when
the
21 question
used to come out as do you recommend
22
approval? And I was very happy to
see today's
344
1
questions not even come close to that sort of
2
working. So, in fact, the
question actually asks
3
only does this provide evidence of efficacy in this
4
subgroup, meaning it could be used for approval, or
5 it
could be used for supportive data perhaps if the
6
company came back with preliminary response rates
7 in
the current ongoing study as opposed to not
8
approval or approval. So, I don't
want anyone on
9
this committee to think that we are going to kill
10 the
drug. Whether we say one thing or
another, it
11 is
simply to provide our opinion about whether or
12 not
the evidence provided today actually shows
13
there is any efficacy.
14
DR. PAZDUR: Donna, the way we
wrote the
15
question specifically--obviously, everything is a
16
risk-benefit decision here. The
efficacy question
17 is
first and, obviously, if that is answered in the
18
affirmative then to go down to look at the toxicity
19
issue.
20
DR. TEMPLE: Actually, you were
putting
21
words in my mouth. I just do want
to say
22
something, I realize people who live in the world
345
1
can't help but think about the implications and
2
what happens if we do this and what happens if we
3 don't.
But we are really supposed to think mostly
4
about whether the therapy shows evidence of
5
effectiveness and not so much about whether people
6
will apply it more broadly than they should and use
7 it
off-label. It is not that we don't ever
worry
8
about that but we are really asking you to focus
9
mostly on whether there is evidence of
10
effectiveness. You know, the
survival of companies
11 is
obviously of interest and whether people become
12
depressed is also of interest but the main thing we
13
need to do and we need your help with is figuring
14 out
whether there is actual evidence of
15
effectiveness for this drug for what they claim.
16
DR. PRZEPIORKA: And having said
that, I
17
would just throw my two cents back in again and
18
indicate that I was impressed with the fact that
19
there are two trials, albeit not perfectly well
20
designed but two trials with very similar results
21 in
terms of the magnitude and the direction of the
22
effect, and most strikingly, similar results with
346
1
regard to outcome. It is very
rare to see two
2
trials, one right after the other, to have the same
3
median survival in both the control group and the
4
experimental group. I thought
that was remarkable.
5 Dr.
D'Agostino?
6
DR. D'AGOSTINO: Again, the first
study
7 had
18 breast cancer patients in it. Really
as a
8
direction it didn't seem to inform the second
9
study. So, retrospectively it is
kind of
10
interesting but prospectively it didn't inform the
11
study at all, and I think it is saying that the
12
third study they are running is exciting. What I
13
tried to say at the end of my earlier spiel is
14
forget the new trial--I have sympathy and am
15
excited about it, but based on the data I think
16
that there are too many questions with the post hoc
17 aspect
of this in the subset that wasn't
18
prespecified for us to give a positive to this
19
first question.
20
DR. PRZEPIORKA: Dr. Buckner?
21
DR. BUCKNER: I also have some
questions
22
about the efficacy issue per se from the data as
347
1
presented as far as response goes.
There were some
2
problems with the methodology in that there was not
3 a
control for dexamethasone. More than 10
percent
4 of
the scans were missing and, of the missing
5
scans, the survival went in favor of the control
6 arm
rather than the experimental arm. The
issue of
7 no
requirement for confirmed response perhaps could
8 be
argued but it doesn't strengthen the data on
9
response. Furthermore, if we are
really looking at
10 the
effect in the brain it would have been very
11
reassuring to have some signal that people were
12
living better with their brain disease in terms of
13
progression either on clinical basis or radiologic
14
basis, and we didn't see that, or some sense that
15 the
death rate from brain metastases was reduced.
16 We
didn't see that either. And, depending
on how
17 you
interpret the quality of life data, the
18
patient-reported data didn't necessarily seem to
19
indicate strong evidence of benefit in the brain
20
either. So, it is always a little
unsettling when
21
endpoints go in opposite directions and that is
22
what I think we have here--I shouldn't say in
348
1
opposite directions but when one endpoint is not
2
supported by multiple other endpoints.
3
DR. PRZEPIORKA: Other comments
from the
4
committee before we call the question?
Dr.
5
Grillo-Lopez?
6
DR. GRILLO-LOPEZ: I have a
general
7
comment about statistics and clinical research, a
8
comment that applies not only to this particular
9
discussion but perhaps to this morning's discussion
10 and
other discussions. As I look at the
membership
11 of
this committee, I see that most of us are
12
clinicians and most of us have been or are
13
currently involved in the care of cancer patients.
14 If
the FDA had been interested exclusively in the
15
statistics behind a clinical trial they would have
16
only statisticians around this table but, in fact,
17 the
majority are clinicians.
18
I think the message the FDA is
giving us
19 is
that they are interested in clinical input, in
20 the
input of those who are actually taking care of
21
these patients and who can, yes, consider the
22
statistics but perhaps consider those statistics as
349
1 a
tool in the decision-making process, a process
2
that also involves making clinical decisions based
3 not
necessarily on numerical or mathematical
4 computations.
5
I think that today, particularly this
6
morning, we have seen the extreme, very eloquently
7
presented, that statistics can go to.
Yes, it is
8 not
that we should ignore statistics but I think
9
there is a limit to how much statistical analysis
10 we
can do and how complex that analysis can become
11
because statistics is a science; it is based on
12
numbers, it is based on mathematics.
Clinical
13
research is an art. It is based
on patients and
14
what happens to patients. And the
more complex the
15
statistical analysis, the more distant you get from
16 the
reality of clinical research, from the reality
17 of
what is happening to patients.
18
So, again, in making our decisions, in
19
making or recommendations to the FDA on these
20
issues we put the statistical analysis on the
21
balance, the results of that analysis on one side
22 of
the balance but we also have to put our own
350
1
clinical opinion of the data and weigh that equally
2 or
perhaps even more strongly than what the numbers
3
alone may say.
4
DR. PRZEPIORKA: Dr. D'Agostino?
5 DR. D'AGOSTINO: I thought this was a case
6
where the statistical issues were quite simple
7
actually. If they had declared
that subgroup
8
breast cancer as the primary group and had given
9 the
right allocation of p values, I think all our
10
votes would be positive. They
didn't do it so it
11 is
not really a complex statistics issue; it is a
12
very simple statistics issue. It
is an unfortunate
13
thing. It may be a real result
but because it was
14 unspecified
and because it was found only in a post
15 hoc
manner we have no way of judging it
16
statistically and I am impressed that you feel you
17 can
judge it clinically without some sort of
18
numerical basis, but that is your prerogative.
19
DR. PRZEPIORKA: Dr. Williams?
20
DR. GRILLO-LOPEZ: I said I was
speaking
21 in
general.
22
DR. WILLIAMS: Donna, I just
wanted to
351
1
clarify. You mentioned that the
question was
2
asking for evidence of efficacy.
Substantial
3
evidence I think is an important term.
It doesn't
4
just mean some evidence, it means enough evidence
5 to
approve it really. That is the term that
is
6
used in the regulation for approval, given that it
7 is
safe enough.
8
DR. PRZEPIORKA: Dr. Pazdur?
9
DR. PAZDUR: I wanted to address
the
10
decision-making process here because I have spent
11
some time on this in my introductory comments.
12
Here, again, we do have statisticians here, we do
13
have clinicians, we have patients and everybody's
14
voice is important. But there is
an underlying
15
process that is unifying decision-making process
16
that all of you must come to.
17
Number one, is there an effect and is it
18
adequately characterized? Number
two, and you can
19
only answer this question if number one is
20
answered, and that is the clinical relevance. But
21 you
cannot make an inference of clinical relevance
22 if
you don't know what you are talking about or if
352
1 it
is poorly characterized. It has to be
there and
2
that is how statisticians help us in making these
3
decisions, especially in a randomized study.
4
Again, remember, this was a randomized
5
study with a primary endpoint of survival with a
6
population that was defined and basically we are
7
looking at subpopulations that were not
8
prespecified.
9
DR. PRZEPIORKA: Any other
comments from
10 the
committee? Dr. Bukowski?
11
DR. BUKOWSKI: I would like to
echo those
12
comments. I think this was a
well-designed and
13
conducted study with predetermined endpoints that,
14
unfortunately, were not met. I
got a little bit
15
confused between eligible and intent-to-treat
16
populations but, notwithstanding, I think the
17
results pretty much hold up. When
you start to try
18 to
define clinical effect and forget the analyses
19
that were presented I think it becomes an issue.
20
Yes, there were two positive studies showing an
21
effect in breast cancer but the way the data was
22
obtained is less than optimal.
So, I am concerned
353
1 by
the findings and their importance. I
think we
2
certainly have to agree that r may well be an
3
effect here but the data speak for themselves.
4
DR. PRZEPIORKA: Further comments
before I
5
call the question? Dr. Reaman?
6
DR. REAMAN: I just want to
respond to Dr.
7 Grillo-Lopez's
statement since he characterized the
8
committee as predominantly clinicians and that we
9 are
to sanction clinical research as an art rather
10
than a science, and I, as a member of the
11
committee, don't believe that we are here judging
12 the
arm of clinical research; it is science.
13
DR. PRZEPIORKA: I think everyone
on the
14
committee would agree with you but thank you for
15
saying that. Other comments? If not, let's go to
16 the
first question, the survival analysis in the
17
overall population was negative.
Do the observed
18
survival results from this single study in the
19
subgroup of patients with breast cancer metastatic
20 to
the brain represent substantial evidence of
21
RSR13 efficacy in this subgroup?
22
Let's start with Dr. Carpenter, please.
354
1
DR. CARPENTER: No.
2
MS. HAYLOCK: No.
3
DR. GEORGE: No.
4
DR. CHESON: No.
5
DR. DOROSHOW: No.
6
DR. RODRIGUEZ: No.
7
DR. PRZEPIORKA: Yes.
8
DR. REDMAN: No.
9
DR. REAMAN: No.
10
DR. TAYLOR: No.
11
DR. MARTINO: No.
12
DR. BUCKNER: No.
13
DR. BUKOWSKI: No.
14
DR. D'AGOSTINO: No.
15
DR. HUSSAIN: No.
16
DR. MORTIMER: No.
17
MS. COMPAGNI-PORTIS: No.
18
DR. PRZEPIORKA: One yes, 16
no. You have
19
your answer and you don't want us to discuss the
20
second question. Any other
information that you
21
want from us?
22
DR. PAZDUR: No.
355
1
DR. PRZEPIORKA: Thank you very
much. I
2
call this meeting adjourned and thank you to all
3 the
committee members.
4
[Whereupon, the proceedings were
5
adjourned.]
6 - - -