View/download the full transcript from Day 2.
October 30, 2017 | Author: Anonymous | Category: N/A
Short Description
.. kind of a double-edge sword for us because it meant. 8. Janet Watkins IN THE CIRCUIT COURT ......
Description
1
1 2 3
U.S. FOOD AND DRUG ADMINISTRATION (FDA)
4
and
5
AMERICAN ASSOCIATION FOR CANCER RESEARCH (AACR)
6 7
PUBLIC WORKSHOP
8 9
Dose-Finding of Small Molecule Oncology Drugs
10 11 12 13 14 15
Tuesday, March 19, 2015
16
8:01 a.m. to 4:05 p.m.
17 18
Washington Court Hotel
19
Washington, DC
20 21 22
A Matter of Record (301) 890-4188
2
C O N T E N T S
1 2
AGENDA ITEM
PAGE
3
Welcome and Work Objectives
4
Amy McKee, MD
5
5
Pasi Janne, MD, PhD
5
6
SESSION 3: Dose-Exposure Exploration
7
Dose-Finding Studies
8 9 10
Qi Liu, PhD
6
Dose Optimization: Ceritinib Dan Howard, PhD
11
Dose Optimization: Axitinib as a Case
12
Example for Dose Titration
13
Yazdi Pithavala, PhD
14
Dose Optimization: Vandetanib
15
Eric Masson, PharmD
16 17 18
42
70
Dose Optimization: Combinations Jin Jin, PhD
88
Moderated Panel Discussion
19
Qi Kiu, PhD
20
Julie Bullock, PharmD
21
17
Audience Q&A
116
131
22
A Matter of Record (301) 890-4188
3
C O N T E N T S (continued)
1 2
AGENDA ITEM
3
SESSION 4: Integrating Dose Optimization in
4
Clinical Development
5
Integrative Approach to Dose Finding
6
PAGE
Geoffrey Kim, MD
7
Feasibility of an Integrative, Adaptive
8
Dose-Finding Trial
9
Rajeshwari Sridhara, PhD
10
Barriers to Implementing Integrative,
11
Adaptive Dose-Finding Trials
12
Lilli Petruzzelli, PhD
13
Practical Considerations of Implementing
14
Dose Optimization Strategies in the Clinic
15
Alice Shaw, MD, PhD
16
Approaches to Integrative, Adaptive
17
Dose-Finding Combination Studies
18
Pasi Janne, MD, PhD
19 20 21 22
A Matter of Record (301) 890-4188
177
189
201
221
253
4
C O N T E N T S (continued)
1 2
AGENDA ITEM
PAGE
3
Future Considerations and Moving Forward
4
Geoffrey Kim, MD
5
Alice Shaw, MD, PhD
275
6
Audience Q&A
291
7
Wrap Up and Adjourn
333
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
A Matter of Record (301) 890-4188
5
P R O C E E D I N G S
1
(8:01 a.m.)
2
Welcome and Workshop Objectives
3
DR. McKEE:
4
Good morning.
Welcome back to
5
our second day of this two-day workshop.
I think
6
we had a very productive day yesterday, and we're
7
hoping that the conversation continues today.
8
for people who were not here yesterday, especially
9
for the panel discussion, we hope to make it as
And
10
interactive as possible within the panel and also
11
including the audience. The one thing we'd like to say is that AACR
12 13
is recording everything, and a transcript will be
14
available after the meeting as well as any slides
15
that are released by the speakers on their website.
16
So if you need any more information about that,
17
please speak to one of the AACR representatives
18
here.
19
DR. JANNE:
Again, I just wanted to welcome
20
everyone for the second day; agreed that we had a
21
very good first day and a good discussion, and hope
22
to continue that same trend today and try to make
A Matter of Record (301) 890-4188
6
1 2 3 4
it as interactive as possible. With that, we'll start our second day, and we'll have our moderators come up here. DR. McKEE:
Our first speaker and moderator
5
for this session is Qi Liu.
6
clinical pharmacology team leaders at FDA. Presentation - Qi Liu
7 8 9
She is one of the
DR. LIU: to Session 3.
Good morning, everyone.
Welcome
As Amy mentioned, my name is Qi Liu.
10
I'm a clin pharm team leader at FDA.
11
to be the moderator for this session, together with
12
my friend and former FDA colleague, Dr. Julie
13
Bullock.
14
dose and exposure exploration.
15
I'm honored
In this session, we'll be focusing on
During this session, I will first give a
16
short 15-minute introduction presentation, and then
17
there will be four presentations, 30 minutes each,
18
from our industry colleagues.
19
going to have one full hour for panel discussion,
20
so we're going to answer all the questions after
21
the presentation is over.
22
After that, we're
The goal of this session is to share our
A Matter of Record (301) 890-4188
7
1
experience on dose selection and to find out how we
2
can do better in the future.
3
from the agency, the aim of my presentation will be
4
to share the lessons that we've learned at FDA.
5
Therefore, since I'm
I'm here to make this presentation with a
6
standard disclaimer that the views in this
7
presentation represent my personal opinion.
8
is truly a collaborative effort.
9
come from the reviews from many FDA folks based on
This
The case studies
10
data from our colleagues from industry and the
11
clinical community.
12
In today's talk, I will first give a brief
13
introduction on the current status of oncology dose
14
selection, and then I will go directly to the eight
15
lessons that we've learned at the agency.
16
use three case studies to illustrate those points.
17
Lastly, I will discuss the goals of Session 3.
18
I will
From year 2011 to 2015, 44 new molecular
19
entities were approved in the Office of Hematology
20
and Oncology Products at FDA.
21
molecule targeted agencies [sic], including
22
16 tyrosine kinase inhibitors.
A Matter of Record (301) 890-4188
Among them, small
For more than half
8
1
of those targeted therapies, the maximum target
2
dose was selected as a recommended dose on the
3
package insert.
4
interest in an alternative approach to the MTD
5
approach, with an aim to optimize effectiveness,
6
minimize toxicity, and to promote patients
7
adherence.
8 9
However, there has been increased
Here are some observations we thought may be helpful to share.
First, like many speakers
10
mentioned yesterday, we felt testing more than one
11
dose in phase 2 or phase 3 trials can be very
12
helpful.
13
exposure-response modeling approach.
14
all available data can be very powerful.
15
We also find a dose-response and Integrating
The target inhibition data can be very
16
informative when the models of action are clear.
17
We've seen that different doses may be needed for
18
different disease settings, even for the same drug.
19
Based on the application we've seen, we also felt
20
that evaluative dose or doses throughout drug
21
development can be very necessary.
22
first case study of nilotinib to illustrate those
A Matter of Record (301) 890-4188
I will use the
9
1 2
points. Another important lesson is that for MTD and
3
DLT determination, the addition to the first cycle
4
data, long-term safety, and tolerability needs to
5
be considered for chronic use of the drugs.
6
my personal favorite points I guess is also
7
Dr. Mark Ratain's favorite point, that we need to
8
appropriately address the interaction between food
9
and drug.
One of
For these two points, I would like to
10
talk a little bit on the case studies of olaparib
11
and ceritinib.
12
Last but not least, allowing up and down
13
titration may be very helpful; however, I'm not
14
going to present any case studies here, as it will
15
be covered very nicely by Dr. Yazdi Pithavala's
16
talk on axitinib.
17
The first case study I would like to discuss
18
is the dose-selection story for nilotinib.
This is
19
a kinase inhibitor indicated for chronic phase and
20
accelerated phase, Philadelphia chromosome-positive
21
CML in adult patients resistant to or intolerant to
22
imatinib.
This drug was initially approved in the
A Matter of Record (301) 890-4188
10
1
U.S. in 2007 for the second indication and was a
2
recommended dose of 400-milligram BID.
3
efficacy supplement was approved for the frontline
4
indication with a recommended dose of 300-milligram
5
BID.
6
In 2010, an
Here's how the dose was selected in the
7
original NDA for the second indication.
In the
8
phase 1 study, the MTD of nilotinib was determined
9
to be 600-milligram BID based on a modified
10
continuous reassessment method.
11
sponsor selected 400-milligram BID dose for the
12
phase 2 study based on safety, PK, and preliminary
13
efficacy data.
14
However, the
In the phase 1 part, the sponsor found the
15
safety of the 400-milligram BID dose was better
16
than 600-milligram BID dose, and it might increase
17
in the systemic exposure of nilotinib when the dose
18
increased from 400 to 600 milligrams
19
The phase 1 dose response and concentration
20
response analysis suggested that higher systemic
21
exposure was related to better efficacy.
22
addition, the sponsor also found the trough
A Matter of Record (301) 890-4188
In
11
1
concentration in the 400-milligram BID is in the
2
concentration range predicted to inhabit targets
3
based on the nonclinical data. Based on the rationales we've just
4 5
discussed, the sponsor went ahead and conducted the
6
phase 2 trial using the 400-milligram BID dose for
7
the second indication, and the trial results and
8
overall benefit-risk assessment supported approval
9
of 400-milligram BID dose in this disease setting.
10
However, risk of QT prolongation and sudden deaths
11
were identified based on the clinical trial data
12
and resulted in a boxed warning on the package
13
insert.
14
The risk of QT prolongation is exposure
15
dependent, and this plot here shows the nilotinib
16
concentration and the QT prolongation risk were
17
positively correlated, and this was one of the
18
major safety concerns for nilotinib at its initial
19
approval.
20
When the sponsor moved forward to the
21
phase 3 trial for the frontline CML setting, they
22
carefully selected dose based on all available
A Matter of Record (301) 890-4188
12
1
data.
Two nilotinib doses were selected for the
2
registration trial for the frontline indication.
3
One of them was 400-milligram BID dose as it was
4
approved for the second indication. The 300-milligram BID dose was selected as
5 6
another dose to test with the hope of reasonably
7
maintaining efficacy while improving the safety
8
based on the exposure-response relationship
9
identified in the second CML setting.
The phase
10
3 data supported 300-milligram BID for this
11
indication.
12
better safety profile compared to the 400-milligram
13
BID dose.
This dose had a similar efficacy but a
14
Here are some important things we learned
15
from nilotinib, and for each point, I also listed
16
some other examples so you can look into them
17
later.
18
information for the drugs on the FDA website.
For most of the examples, you can find more
19
For nilotinib, the sponsor tested two doses
20
for their frontline CML trial, which helped them to
21
find a lower dose to maintain the efficacy but
22
improved the safety profile.
The sponsor used both
A Matter of Record (301) 890-4188
13
1
the dose-response and exposure-response modeling
2
approach to make their dose selection.
3
used the target inhibition data to justify their
4
dose selection.
They also
They identified different doses for the
5 6
front line and second line in order to optimize the
7
benefit-risk profile.
8
selection throughout drug development.
9
continuous paradigm paid off in the end.
They evaluated the dose This
The second case I would like to share is for
10 11
olaparib.
12
gBRCA mutated advanced ovarian cancer after 3 or
13
more prior lines of chemo.
14
submitted, the original indication as proposed by
15
the applicant was for maintenance treatment for
16
platinum-sensitive relapsed gBRCA mutated ovarian
17
cancer.
18
from the ODAC.
19
this unfavorable vote was the toxicity of the
20
therapy and the risk of MDS and AML for patients
21
not otherwise undergoing treatment.
22
This is a PARP inhibitor approved for
When the NDA was
However, this received an unfavorable vote And one of the major reasons for
In this case, I would like to talk a little
A Matter of Record (301) 890-4188
14
1
bit about dosing with regard to food and its
2
impact.
3
members expressed concern over the prolonged GI
4
tolerability issue in a maintenance setting.
5
the clinical trial, patients received BRCA mutated
6
ovarian cancer who received 3 or more prior lines
7
of chemo.
8
43 percent had vomiting, and 31 percent had
9
diarrhea.
During the ODAC discussion, many panel
In
Sixty-four percent of them had nausea,
10
In the clinical trials, olaparib was taken
11
under fasting condition, and this drug needs to be
12
given twice a day.
13
effect data was submitted and issued that fruit had
14
minimum impact on olaparib PK.
15
recommended allowing the drug to be taken without
16
regard to food for patients' convenience and
17
potential improvement in the GI tolerability of
18
this drug.
19
During the NDA review, food
Therefore, we
The last case I would like to share is for
20
ceritinib.
This is a kinase inhibitor approved for
21
ALK positive, metastatic, non-small cell lung
22
cancer, progressed on or intolerance to crizotinib.
A Matter of Record (301) 890-4188
15
1
The approved dose for this drug is 750 milligram
2
once daily on an empty stomach, that is do not
3
administer the drug within 2 hours of a meal.
4
One issue with this drug is severe
5
persistent GI toxicity in the registration trial
6
dose modification due to GI tox occurred in 38
7
percent of the patients.
8
trial, we find that food can increase ceritinib
9
exposure.
10 11
From the food effect
A low-fat meal or a high-fat meal could
increase AUC by about 60 to 70 percent. At a scientific conference meeting, an
12
investigator associated with the registration trial
13
talked about recommending food intake with the drug
14
administration to improve GI tolerability and
15
compliance.
16
requested a postmarketing trial to compare the
17
systemic exposure and safety of 750 milligram in
18
the fasted state and the lower dose with food.
19
aim of the study is to see whether we can give a
20
reduced dose of ceritinib with food to improve the
21
GI tolerability in patient compliance.
22
FDA discussed with the sponsor and
The
From the olaparib and ceritinib case, we can
A Matter of Record (301) 890-4188
16
1
see that in addition to the first cycle safety and
2
tolerability, long-term safety and tolerability
3
needs to be considered for chronic use of the
4
drugs.
5
importance of evaluation and the proper strategy to
6
handling food and drug interaction.
7
From both cases, we can also see the
Here's some food for thought.
What if we
8
study the food effect early on and use this
9
information to help dosing strategy in the
10
registration trials?
11
benefit-risk profile of the drug and patient
12
compliance?
13
Will this improve the
In conclusion, careful balance of benefit
14
and risk is critical in oncology dose selection,
15
and improvement can be made to traditional MTD
16
approach to optimize effectiveness, minimize
17
toxicity, and to promote patient adherence.
18
The goal of this session is to share dose
19
selection and optimization experience among the
20
industry, agency, and the clinical community and to
21
discuss the different dose-finding approaches,
22
their successes, limitations, and challenges, and
A Matter of Record (301) 890-4188
17
1
to discuss tools that can be utilized, including
2
their strengths and limitations.
3
know what we can do better in the future.
We also want to
With this, I'm going to close my talk, and
4 5
our next speaker is Dr. Dan Howard.
6
global head of oncology and clinical pharmacology
7
at Novartis.
8
developer story on the dose optimization for
9
ceritinib.
Dan is going to tell us the drug
10
Dan, the podium is now yours.
11
(Applause.) Presentation - Dan Howard
12
DR. HOWARD:
13 14
He's the
Welcome to day 2 of the
ceritinib story.
15
(Laughter.)
16
DR. HOWARD:
My name is Dan Howard.
Thank
17
you very much for the introduction.
18
very much to the organizers for inviting me to
19
present the story of the development of ceritinib.
20
I have a few things that I need to share with you
21
first.
22
And thank you
I only joined oncology about three years
A Matter of Record (301) 890-4188
18
1
ago.
2
years in drug development without ever working with
3
an oncology drug.
4
wanted to be a part of an area where I could see a
5
lot of excitement was going on, but I also wanted
6
to learn a lot about how drugs were developed in
7
this area.
8 9
I joined oncology after working for nearly 20
I joined oncology because I
When I joined in 2012, our early development group was working on a new drug.
We called it
10
affectionately LDK378.
It became known later on as
11
ceritinib.
12
clinic -- we had been working on the drug in the
13
clinic for a little more than a year.
14
of weeks into the new position that I had, I got a
15
phone call from our scientists saying, "Dan, we
16
need you to assign some additional resources to the
17
development of this drug.
18
exciting things happening."
We worked on this drug in the
In a couple
We've got some really
19
I was told that the first-in-human data
20
indicated that we saw a response rate of nearly
21
60 percent in anaplastic lymphoma kinase positive
22
cancer patients that were resistant to crizotinib.
A Matter of Record (301) 890-4188
19
1
I remember my first thought when I heard that.
2
thought, "What is crizotinib --"
3
(Laughter.)
4
DR. HOWARD: -- "and do we own it."
5
(Laughter.)
6
DR. HOWARD:
I
And I owe a big thanks to the
7
scientists who did the work on this.
I was not
8
involved in the early development.
9
involved, as I said, at the point where we began to
I became
10
realize that we had a potential candidate here for
11
treating ALK-positive lung cancer patients.
12
What I'm going to present to you
13
today -- none of the slides I'm going to present
14
are new.
15
places.
16
slightly different perspective, a perspective of
17
somebody who, although an insider, was definitely
18
an outsider from the drug development of oncology.
19
All of these have been presented in other But what I will present to you today is a
Now, I quickly learned a few important
20
things about crizotinib and ceritinib.
21
learned that crizotinib was not our product --
22
(Laughter.)
A Matter of Record (301) 890-4188
First, I
20
DR. HOWARD:
1
-- and that was a bummer.
But
2
I also found out that these two products shared the
3
same mechanism of action.
4
ALK-positive inhibition, or -- excuse me, the
5
anaplastic lymphoma kinase inhibition that
6
ultimately led to the apoptosis of the echinoderm
7
microtubule associated protein like for ALK fusion
8
gene or the EML for ALK transfusion, or
9
translocation gene.
Namely, it was the
This gene has been associated with anywhere
10 11
between 5 and 7 percent of all non-small cell lung
12
cancers, and because the translocation was only
13
associated with abnormal cells -- in other words,
14
it was not associated with normal cells -- we had
15
good reason to believe that there would be an
16
expectation of a high therapeutic index for the
17
drug.
18
In addition, we knew that, from our
19
nonclinical data, ceritinib looked to be about 20
20
times more potent in this activity than crizotinib.
21
We also knew that we had a slightly longer half-
22
life, and so we were anticipating that we might be
A Matter of Record (301) 890-4188
21
1
able to give the drug less frequently.
2
knew that our product had the ability, potentially,
3
to provide for crossing the blood-brain barrier,
4
and therefore could be beneficial, possibly, in
5
metastatic brain cancer.
6
important things to keep in mind as I move forward
7
in the presentation.
Now, these were all
Our first-in-man trial was designed to
8 9
We also
characterize, as most are, the pharmacokinetics
10
tolerance and early activity of the drug.
11
you'll recognize the design as pretty conventional
12
as a dose-escalation design.
13
escalated the doses until we would reach MTD.
14
would use the information, then to select the
15
recommended phase 2 dose by expanding into the dose
16
that we felt was the likely dose to move forward
17
with.
18
Now,
We basically We
Of the ALK positive in this trial, we
19
allowed basically any cancer which was ALK
20
positive.
21
ultimately entered into the trial, only 14 of them
22
presented with cancers other than non-small cell
Of the 304 patients, however, that were
A Matter of Record (301) 890-4188
22
1
lung cancer. All patients were fasted.
2
Ceritinib has a
3
low permeability and a low solubility based on our
4
preclinical experiments.
5
on some modeling, that there would be an increase
6
with food administration.
And we did expect, based
The formulation that we used was the kind of
7 8
formulation we generally take into our phase 1
9
trials.
It was a very low-complexity, simple
10
formulation.
It was manufactured, in fact, at
11
three different sites in order to provide enough to
12
move forward.
13
be the final formulation that we would take forward
14
to the marketplace.
It was not initially anticipated to
Now, one should take note that there are 59
15 16
patients in the dose-escalation phase, and of
17
those, 10 received daily doses between 50 and
18
300 milligrams; 39 received doses between 400 and
19
700; and 10 in the dose escalation part received
20
750.
21 22
One of the key objectives was to determine the MTD.
Now, between 50 and 750 milligrams, the
A Matter of Record (301) 890-4188
23
1
drug was well tolerated with a principal side
2
effect that we observed being GI.
3
750-milligram dose with a predicted probability of
4
an overdose that was well below that for
5
intolerability, which we had listed as 33 percent.
6
We reached the
The next dose for escalation would have been
7
900 milligrams.
However, due to the frequency of
8
persistent grade 1 and 2 GI toxicity and an
9
occasional grade 3 to 4 liver enzyme increase,
10
elevation, at 400 milligrams and above, the
11
investigators working with us on the trial elected
12
not to go further and test a 900-milligram dose.
13
The safety of the 750-milligram dose was
14
confirmed in the expansion patients.
In fact, we
15
expanded first into a group of 10 subjects to get a
16
further sense of the safety.
17
additional group, in the first group, there were
18
zero DLTs out of 10.
19
were 2, which was only 11 percent, which was
20
definitely well below what we would regard as
21
intolerable.
22
intolerable dose, we declared the 750 as the
And in that
In the second group, there
And though we did not reach an
A Matter of Record (301) 890-4188
24
1
maximum tolerated dose and examined the data for
2
the additional information, which would provide
3
insight into dose escalation.
4
Now, it's important -- I want to point out
5
something important here.
6
achieve MTD.
7
inside of a tolerable limit.
8
higher dose.
9
but we had never actually achieved MTD.
10
We didn't actually
We found the 750-milligram dose was We did not go to a
We had a maximum administered dose,
The drug was clearly showing effectiveness
11
at doses 400 milligrams and above, and of the 114
12
patients with the ALK-positive, non-small cell lung
13
cancer, we had an overall response rate of about
14
58 percent.
15
response either for efficacy across the change in
16
tumor size, across the 400 to 750 dose range, or
17
for safety for that matter.
18
an exposure-response difference between the GI
19
effects that we were observing across the 400 to
20
750 dose range, or for that matter, the liver
21
enzyme increases.
22
We observed that there was no exposure
We didn't actually see
It wasn't so much a surprise that we didn't
A Matter of Record (301) 890-4188
25
1
see it for the GI.
2
due to exposure; it was probably due to local.
3
the one hand, we had expected a very high
4
therapeutic index, and the lack of an exposure
5
response with the GI activity really just suggested
6
to us that it was local.
7
The GI toxicity probably wasn't On
Our preclinical models clearly indicated
8
that the greatest tumor inhibition and duration of
9
response was associated with the highest exposures,
10
in fact, higher than the exposures that we were
11
able to achieve with the 750-milligram dose.
12
fact, according to our preclinical model, we needed
13
about twice as much exposure as we were able to
14
achieve.
15
In
We knew, based on this preclinical data,
16
that we needed to push the dose if we wanted to
17
achieve the kinds of duration of response, the
18
kinds of response we wanted to get in patients.
19
also knew that at the 750-milligram dose, exposure
20
in humans exceeded the highly active level of what
21
we see here as 50 milligrams per kilogram, but did
22
not exceed the exposures required in the
A Matter of Record (301) 890-4188
We
26
1
crizotinib-resistant models. In the crizotinib-resistant, ALK-positive,
2 3
non-small cell lung cancer models, the higher doses
4
of LDK378 were required for tumor regression than
5
we could achieve even with the wild-type kinase
6
domain.
7
our patients were still on ceritinib, which was
8
kind of a double-edge sword for us because it meant
9
that our data were not robust enough -- it meant
And at the time of dose selection, many of
10
that we should be very excited about what we were
11
seeing, but our data would never be robust enough
12
for us to estimate a duration of response.
13
were not able to take that into account when we
14
were selecting the doses.
15
So we
In May of 2012, we debated the selection of
16
dose, and we concentrated on the data that was
17
contained in the range of 500 milligrams and above.
18
It represented about 30 patients worth of data at
19
the time.
20
the start of a food effect study in healthy
21
volunteers.
22
the phase 2 study that we would ultimately plan to
We in clinical pharmacology accelerated
And in May, we debated the designs of
A Matter of Record (301) 890-4188
27
1
conduct.
But in June, we realized that we might be
2
able to convert our phase 1 study over the
3
registrational status by expanding in a larger
4
number of subjects at the defined dose. In July then, we modified this study to
5 6
increase the number of subjects in the expansion
7
and phase and reduce the interval between the last
8
crizotinib treatment and the first ceritinib
9
treatment, and we facilitated the collection well,
10
so facilitated the collection of imaging data for
11
independent review by a CRO. In November, we met with the agency and
12 13
discussed our plan to convert this phase 1 study
14
over to a registrational study, and the agency
15
agreed.
16
preparing for all of these changes, something
17
happened in March.
18
therapy designation.
19
the time that our product had gone from a
20
research-based, liberal kind of let's understand
21
what we're doing, let's learn more about what's
22
going on.
And then something happened.
As we were
We were given breakthrough And I joked with the team at
It had become a very conservative
A Matter of Record (301) 890-4188
28
1
Republican at this point.
2
(Laughter.)
3
DR. HOWARD:
And it did not want anyone to
4
mess with that trial design.
5
anybody to mess with what we were we doing on the
6
study.
7
confirm that our selected dose was the right dose.
8 9
It did not want
It just needed us to collect information to
At one point, at least in our company, we were elated.
At the time, in March of 2013, we
10
were only the third breakthrough therapy
11
designation that had been given at the time, and
12
the company was just ecstatic.
13
though, we were scared to death because suddenly
14
now, the timelines for the project were the
15
timelines for this phase 1 study.
16
squeeze everything we needed to learn into the span
17
of the timeline for that one study.
18
In clin pharm,
We needed the
Everything about the development of this
19
compound became high priority.
20
study, our phase 1 study was no longer about
21
learning; it was now about confirming.
22
a study to further optimize the dose regimen, and
A Matter of Record (301) 890-4188
As a registrational
It was not
29
1
it's timeline, as I said, became a timeline for the
2
entire project.
3
was acceptable, and fortunately our capsule
4
formulation we learned could be scaled to
5
commercial manufacturing.
6
Fortunately, our safety profile
We confirmed that there was a positive food
7
effect in a separate PK study of healthy
8
volunteers, but there was no way for us to
9
incorporate this knowledge into our ongoing trial
10
without extending the duration of the trial or
11
without expanding the number of subjects.
12
Now, mind you, those subjects we were
13
recruiting were not easy subjects to recruit.
14
These subjects are only 5 to 7 percent of all non-
15
small cell lung cancer patients.
16
would require us to add an additional arm of
17
another 200 subjects to demonstrate an alternative
18
dosing regimen was not going to be favorably looked
19
upon by either our investigators, our patients, or
20
for that matter, our company.
21 22
Anything that
Our clinical pharmacology department immediately implemented our mass balance study,
A Matter of Record (301) 890-4188
30
1
feeling that it was important, and some key drug
2
interaction studies that we knew we needed to get
3
incorporated into the development so we would be
4
ready for submission.
5
Once we completed the expansion cohort, our
6
initial assessments of response that we saw in the
7
first 59 patients were validated.
8
seeing an overall response rate that was very
9
similar to what we observed in the first 59, about
10
We ended up
55 percent in the pretreated patients.
11
Our safety profile primarily consisted,
12
again, of the liver enzyme increases and the GI
13
effects that we saw.
14
most often grade 2, and they occurred in
15
about -- well, actually, in more than 80 percent of
16
the patients who were treated.
17
In fact, the GI effects were
Dose reductions and interruptions occurred
18
in 59 percent of the patients treated, but it
19
helped patients to remain on therapy.
20
overall dropout rate, discontinuation rate of
21
patients due to AEs was only 10 percent.
22
clearly, the modification of the dose allowed us to
A Matter of Record (301) 890-4188
In fact, the
So
31
1
maintain the treatment even though we had an
2
80 percent observation of this effect.
3
Here we see the logistic regression for the
4
probability of response by exposure quartiles for
5
ceritinib expressed as an average steady-state
6
trough concentration.
7
what you're going to see is that there's very
8
little change in the probability of response.
9
Across the exposure range,
Now, something we should keep in mind -- and
10
I'm sure many of you have already been here
11
before -- this data is predominantly one dose.
12
It's the exposure driven almost exclusively by 750
13
milligrams because that's where 245 of our patients
14
existed.
15
The FDA analysis for doses -- similar
16
analysis for efficacy, telling the difference
17
between our doses -- divided it into two groups:
18
50 to 700 milligrams and a second group of
19
750 milligram.
20
the overall response rate or in the duration of
21
response.
22
only 44 patients in the 50 to 700-milligram group,
And they showed no difference in
Now, keep in mind of course, there are
A Matter of Record (301) 890-4188
32
1
and there are 246 actually in this set for the
2
750-milligram group.
3
you see here is that there's very little, if any,
4
difference we can see from an exposure response.
5
Despite that difference, what
What was clear at the higher doses was that
6
they were associated with shorter times to dose
7
reduction or interruption, though not permanent
8
discontinuation.
9
higher exposures, then, were not associated with a
This is interesting because the
10
higher rate of discontinuation.
11
important.
12
Now, this is
This tells us that patients were not leaving
13
the study -- that the patients who were leaving
14
this study were not the patients who were in the
15
highest exposure range.
16
that by adjusting the dose, we can maintain the
17
patients on the trial, but this is not a story like
18
axitinib.
19
It tells us that we know
As previously stated, we were able to
20
measure some exposure response for non-GI related
21
adverse events.
22
C-trough probability curves for probability of an
And what you see here are the
A Matter of Record (301) 890-4188
33
1
event versus the quartile exposure, and in the
2
lower right-hand corner, you see the actual
3
concentration response for QTc.
4
What you can see in this graphic is that the
5
liver enzyme increases, glucose elevation, and all
6
grade 1 and above the GI AEs -- excuse me, for the
7
first two, for the liver enzyme increases and for
8
the glucose elevations, we can see that there is
9
definitely an increase in the probability of having
10 11
that adverse event with an increased exposure. What we see on the lower left-hand side is
12
that, again, as I said before, there's no real
13
connection between the GI AEs and the
14
concentration, as expected.
15
there was an increase in QTc lengthening as a
16
result of increased exposures.
17
the top end was about 17 and a half milliseconds.
But we also see that
That increase at
18
There was no one adverse event that
19
dominated the AE profile or which led to study drug
20
discontinuation.
21
caused people to discontinue were distributed
22
pretty evenly across all commonly seen adverse
The adverse event profile that
A Matter of Record (301) 890-4188
34
1
events in the lung cancer population.
2
frequent reasons for dose reductions and
3
interruptions were the liver enzyme elevations.
4
And of course the AEs due to GI, of course those
5
were also the most dominant AEs that we saw across
6
the entire drug profile.
7
reasons, however, for ceritinib discontinuation.
8
At the 750-milligram dose, the exposures
9
And the most
These were uncommon
required for complete tumor regression, as I said
10
before, were not being achieved as indicated by the
11
preclinical model, and the average area under the
12
curve, in fact, at the 500-milligram dose, matched
13
the area under the curve observed preclinically for
14
90 percent tumor regression.
15
However, the interpatient variability was up
16
around 60 percent, which, as we found in clin
17
pharm, supported the need to go as high as we could
18
to make sure that as many patients were treated
19
with an efficacious exposure as possible.
20
operating with the knowledge that we needed to push
21
the exposure as much as possible to maintain and
22
achieve the efficacy that was indicated by our
A Matter of Record (301) 890-4188
We were
35
1
preclinical models. Several modeling approaches were attempted
2 3
during the course of this study.
What I'm showing
4
you here is an Emax model, an Emin model if you
5
will, to look at, in this case, best tumor-size
6
change across the trough values associated with
7
doses 400 and above.
8
couple of things.
9
obviously very large.
And what you see here is a
One, the variability is
The variability in exposure and the limited
10 11
number of patients in the phase 1 study provided
12
very little ability to discriminate the effect by
13
dose.
14
turned out, in the 6 patients who were given that
15
dose, actually had a higher exposure than they did
16
at the 750, owing to the variability.
17
In fact, the 700-milligram dose, as it
Ultimately, the concentration range overlap
18
for doses 400 and above for the average Cmin meant
19
that the PK/PD modeling, regardless of what we
20
tried, resulted in almost on/off effects for tumor
21
shrinkage.
22
variability and the preclinical evidence supporting
The relatively high interpatient
A Matter of Record (301) 890-4188
36
1
the highest possible dose led us to select the
2
750-milligram dose to avoid underexposing patients
3
inside of this high degree of variability.
4
Now, we knew from our initial data in the
5
phase 1 study, even in the 59 patients to start,
6
that GI tolerance was a concern, and the data from
7
all our subjects who participated in this study at
8
the 750-milligram dose is shown here.
9
before, there was no relationship between the
10
exposure and GI intolerance.
11
clear relationship, either, with dose.
12
I stated
In fact, there was no
Since concomitant administration with food,
13
as we all know, is often associated with a
14
reduction in GI distress, we gave consideration for
15
giving ceritinib with a meal.
16
that if we gave it with a meal, we knew that we
17
would increase the exposure.
18
increasing the exposure could lead to increased
19
adverse events:
20
for instance, QTc.
21 22
However, we knew
We knew that
liver; for instance, hypoglycemia;
Now, we tested -- and what I'm showing here is the low fat and the high-fat meals that were
A Matter of Record (301) 890-4188
37
1
tested with ceritinib to determine what the impact
2
was.
3
58 percent for the low fat and 73 percent for the
4
high fat.
5
standard low-fat meal.
6
calorie/half fat meal by comparison to the standard
7
high-fat meal recommended in the guidance.
8
Incidentally, for those keeping score, GastroPlus
9
did indicate that there would be about a 50 percent
And we observed an increase in exposure of
And just for the record, there is no What we used was a half
10
increase with a high-fat meal, so it was at least
11
in the ballpark.
12
Ceritinib is a BCS-4 compound with both low
13
solubility and low permeability.
14
population PK modeling, we performed Monte Carlo
15
simulations for steady-state exposures to 450 given
16
with a low-fat meal and 600 milligrams given with a
17
low-fat meal, to determine where they would end up
18
relative to the target exposure that we were
19
achieving with the 750-milligram fasted
20
administration condition.
21 22
And using our
What we found was that owing to the fact that it was at steady state, which reduces some of
A Matter of Record (301) 890-4188
38
1
the variability, and owing to the fact that
2
increasing the bioavailability of the drug also
3
decreased a smaller amount, the variability, what
4
we were able to find in that Monte Carlo simulation
5
was that the concentrations, the predicted
6
exposures, ended up in the lower half of what we
7
would call a bioequivalent range for the 450 given
8
with a low-fat meal, and in the upper half for 600
9
milligrams given with a low-fat meal.
10
This was a problem for us because, now, in
11
order for us to make a recommendation about either
12
one of these conditions, we really needed to
13
investigate it.
14
we needed to study whether or not we could maintain
15
the effectiveness of the drug and also, more
16
importantly in this case, not end up with a drug
17
that had a different benefit-risk ratio because of
18
a higher safety problem.
19
We needed to go into patients, and
Actually in discussions with the agency, we
20
agreed that it was necessary to go in and conduct
21
such a study.
22
currently ongoing, where we have 450 milligrams
And in fact, such a study is
A Matter of Record (301) 890-4188
39
1
once daily with a low-fat meal and 600 milligrams
2
once daily with a low-fat meal compared to 750 with
3
the fasted conditions in the label.
4
Now, I can tell you -- and you can read this
5
slide yourself.
I can tell you that what we have
6
done in order to develop ceritinib was get the drug
7
to patients who had no other option when they were
8
unable to be effectively treated by crizotinib.
9
order to get this drug in, we needed to get this
In
10
drug with information that supported a benefit-risk
11
ratio that would allow us to effectively treat
12
patients, and we did that.
13
not only very quickly, but we did it with the most
14
efficient means that we could.
15
We did that I might say
We are in the process, as a breakthrough
16
therapy, of developing further an understanding of
17
the optimization of that dosing regimen.
18
if there's anything to learn from this ceritinib
19
experiment or the ceritinib development, there are
20
five things I really want to share with you that I
21
think are extremely important, as I think -- and
22
many of you are going to say, "I've already been
A Matter of Record (301) 890-4188
I think
40
1
there.
2
where we are."
3
We've done that before.
This is exactly
Well, first of all, if your drug has the
4
potential for breakthrough therapy, use a
5
formulation that has the potential for
6
commercialization.
7
manufacturing facilities needed to produce it.
8
Now, with ceritinib, we were fortunate
9
Limit the number of
because we were able to go to a commercialization
10
with the formulation we took into phase 1, but
11
that's not true of all of our phase 1 studies.
12
if you want to look at how difficult it can be for
13
anyone, look at olaparib as an example of how you
14
can really put yourself into a situation where
15
there's some risk involved.
16
And
We learned that dose and regimen
17
considerations, surprise, should include food.
18
Now, this is especially true when GI tolerance may
19
limit the doses.
20
early, especially when you're looking at GI
21
intolerance as a limiting factor on continued
22
administration of the drug.
It is important to study it
A Matter of Record (301) 890-4188
41
We also emphasized, as a result of this
1 2
study and others as well, that we should be
3
including ICH14 compliant QTc assessments in our
4
phase 1 first-in-human studies regardless of the
5
preclinical results from our telemetry studies.
6
This is because you can be surprised when you do
7
not observe an effect -- in your preclinical
8
studies, you can be surprised by an effect observed
9
in humans. Fourth, using this as an example, we've
10 11
begun to take a serious look at what I am referring
12
to as strategic enhancement of multiple cohorts in
13
our phase 1 studies to provide additional
14
information and improve our exposure-response
15
modeling.
16
maybe we should be expanding into more than one
17
dose.
18
is that, yes, there are occasions where expanding
19
into more than one dose may be helpful.
20
We saw yesterday the recommendation that
I think what ceritinib helped us understand
Fifth, we are encouraged to complete
21
evaluation of all tolerable exposures in phase 1.
22
Now, what am I really saying there?
A Matter of Record (301) 890-4188
As a clinical
42
1
pharmacologist, I truly wish we had gone to
2
900 milligrams.
3
about how far you can go.
4
dose, if it had turned out to be the maximum
5
tolerated dose, it would have provided essential
6
information about our therapeutic index that would
7
be helpful in further development.
8
from my point of view, it helps me to protect
9
patients from overexposure when, for instance,
I understand that there is a limit But with a 900-milligram
And certainly,
10
organ impairment or drug interaction occurs.
11
is not the enemy.
12
default condition for selecting your dose.
13
MTD
The enemy is using MTD as the
With that, I'd like to thank you very much
14
for your attention.
15
(Applause.)
16
DR. LIU:
Our next speaker is Dr. Yazdi
17
Pithavala.
18
pharmacology at Pfizer.
19
presentation is Dose Optimization: Axitinib as a
20
Case Example for Dose Titration.
21 22
He's the senior director of clinical The title of his
Thank you.
Presentation - Yazdi Pithavala DR. PITHAVALA:
Good morning, everyone.
A Matter of Record (301) 890-4188
My
43
1
name is Yazdi Pithavala.
2
pharmacologist at Pfizer.
3
thanking the organizers for allowing me to
4
participate on what for me personally is turning
5
out to be a very relevant and very useful workshop.
6
I'm a clinical I want to start by
Dr. Nitin Mehrotra, at the session
7
yesterday, through his case examples covered
8
axitinib very eloquently.
9
presentation here is to kind of do a deeper dive on
And the goal of my
10
the dose titration principle as we applied it for
11
axitinib during the clinical development program,
12
and kind of go through the specifics of why
13
titration, how titration, when titration, and all
14
the challenges that are involved as well as the
15
benefits associated with it.
16
Axitinib is one of the more potent VEGF
17
receptor antagonists.
It's a second generation
18
inhibitor.
19
treatment of second-line RCC after failure of
20
either a cytokine or a prior VEGF-based therapy.
21
The approval was based on a phase 3 study against
22
an active comparator.
It was approved in 2012 for the
In this case, it was
A Matter of Record (301) 890-4188
44
1
sorafenib.
2
is the axitinib molecule docking fairly tightly
3
into the juxtamembrane domain of the ATP,
4
deep-binding pocket of the VEGF receptor-2 protein.
5
So it's a fairly avidly, tightly-bound compound.
6
And what you see in the pictorial there
This is the outline of my presentation.
I
7
want to say that we decided -- and I hope to
8
convince you by the end of my presentation -- the
9
same; that we decided fairly early on that this was
10
a drug which could benefit from dose titration, and
11
we implemented it fairly early in the clinical
12
development program.
13
One of the downfalls of dose titration is
14
that when we first presented it to people, both
15
internally as well as externally, the first impulse
16
that a lot of people have is that clinical
17
pharmacology hasn't done their job, that we don't
18
know what the dose of the drug is, and we are just
19
saying, eh, you take your patient, figure it out,
20
personalizes the dose, find out what the dose is.
21 22
I want to convince everyone here, that's far from the truth.
We do believe we have the right
A Matter of Record (301) 890-4188
45
1
dose for the majority of patients.
2
there's a subset of patients who stand to benefit
3
from dose titration that we are implementing this
4
for.
5
how we came about picking up the starting dose, or
6
the MTD dose, for this particular compound, and
7
then jump into the retrospective analysis that we
8
performed using data from three phase 2 studies to
9
provide conclusive evidence for why dose titration
10
It's just that
So I wanted to go into the justification for
might be beneficial.
11
Then, in the second part, I want to go over
12
the specifics of dose titration: what was the exact
13
algorithm that we used; when do you implement dose
14
titration; what are the metrics; how many patients
15
end up dose titrating; how does that impact safety;
16
how does that impact efficacy; how does that
17
complicate your drug development program, and then
18
wrap up with a prospective study that we designed
19
to conclusively, once and for all, find out if dose
20
titration was beneficial or not.
21 22
So what was the justification for the 5-milligram dose?
Our first-in-human study was
A Matter of Record (301) 890-4188
46
1
conducted with 36 patients.
It was a standard dose
2
titration, the 3 plus 3 design that we all loved
3
and hated at our session yesterday afternoon.
4
was a standard 3 plus 3 design, and using the
5
standard toxicity criteria, 5-milligram BID was
6
identified as the MTD dose.
It
Axitinib has a fairly short plasma half-life
7 8
of about 2 to 5 hours, so we know that to maintain
9
plasma concentrations continuously, it has to be
10
given twice a day.
11
option in terms of backing down to less frequent
12
dosing.
13
safety perspective, our MTD dose.
14
So we don't really have much
So 5-milligram BID was, at least from a
One of the things we did in our
15
first-in-human study was that we built in a fairly
16
rigorous pharmacodynamic marker for every patient
17
in our first-in-human study.
18
angiogenic drug.
19
flow to tumors, so we had dynamic contrast,
20
enhanced MRI done in every evaluable patient in our
21
first-in-human study.
22
This is an anti-
It's supposed to reduce blood
So across the doses that were tested, we
A Matter of Record (301) 890-4188
47
1
were able to look at changes in blood flow and
2
permeability in selected lesions in all of these
3
patients, and what we saw from those
4
pharmacodynamic studies was that actually
5
5-milligram BID was also supported by results from
6
that study.
7
through those results in a little more detail.
8
The other piece is the efficacy piece.
9
In the next few slides, I will go
Our
very first phase 2 study, which was done in
10
second-line RCC studies, we used a flat dose of 5-
11
milligram BID.
12
5-milligram BID flat dose.
13
particular dose, we saw a fairly robust clinical
14
response, overall response rate of about 44 percent
15
and a PFS of 13.7 months.
Almost every patient got And using that
16
So you don't really need dose titration,
17
even with a 5-milligram BID dose, we are able to
18
get fairly good clinical response.
19
we were greedy.
20
even from our first-in-human study, that there was
21
a substantial intersubject variability, which is
22
not uncommon for most oral drugs.
It's just that
We thought that we recognized,
A Matter of Record (301) 890-4188
And while the
48
1
majority of patients are able to get the
2
appropriate exposures, there is a subset of
3
patients who are getting subtherapeutic exposures.
4
And in those individuals, why not give those
5
individuals an opportunity to dose increase? These are the pharmacodynamic results from
6 7
the first-in-human study.
This is a fairly
8
dramatic response that we saw in the patients who
9
had an adenoid cystic carcinoma in the upper palate
10
of the mouth.
What you can see is functional
11
imaging.
12
flow.
13
of high blood flow, and the blues and the whites
14
represent areas of low blood flow.
Again, what we're looking here is blood
And the reds and the yellows represent areas
So you can see at baseline, this patient had
15 16
a fairly vascular lesion in their mouth.
17
middle areas were necrotic.
18
had a lot of blood flow; fairly rapid.
19
2 days of dosing, you start seeing a reduction in
20
blood flow, which is more enhanced by week 4 and by
21
week 8.
22
The
The surrounding areas After
We are able to then analyze this data and
A Matter of Record (301) 890-4188
49
1
look at what the exposure response is for this
2
particular reduction in blood flow.
3
plotted here is on the X-axis, we have the exposure
4
for axitinib, the 24-hour exposure.
5
have on the Y AUC is the reduction in blood flow.
6
The parameter there is AUC.
7
imaging modality.
8
reduction, the greater in blood flow.
And what is
And what we
It's the AUC for the
And obviously, the greater the
9
What you can see is that fairly early on,
10
you see a very dramatic drug-related effect, that
11
at fairly low exposures, you already start seeing
12
reductions in blood flow.
13
AUC of around 250 to 300, you've kind of max'd out
14
this effect, and you don't see any further
15
reduction in blood flow at higher exposures.
16
But then once you hit an
We were very lucky because this particular
17
exposure is right around the mean AUCs that you see
18
with the 5-milligram BID dose, so the dose which is
19
associated with the saturable effect nicely
20
dovetails with the dose which is determined to be
21
the MTD from the safety evaluations.
22
say we looked at other pharmacodynamic markers such
A Matter of Record (301) 890-4188
I should also
50
1
as circulating VEGF receptor-2, circulating c-Kit,
2
and very similar results.
3
at around the exposures we get with the 5-milligram
4
BID dose.
5
You max out the effect
We implemented dose titration moving forward
6
in our phase 2 studies, and we were able to pool
7
data from three separate second-line, RCC, phase 2
8
studies that were conducted.
9
60 patients.
Each had about
One study was in cytokine refractory
10
patients in the U.S., one was in sorafenib
11
refractory patients in the U.S., and the third
12
study was cytokine refractory patients in Japan,
13
done to support the Japanese registration.
14
We were able to pool the data from those
15
studies, and what we saw is that using a sample
16
population PK model and doing PK/PD analysis, those
17
patients who get higher than the median exposure do
18
better compared to those who don't; fairly standard
19
clinical pharmacology 101.
20
divide this into quartiles, we see the same effect.
21
Very similar response is seen using overall
22
survival as well, that patients will do better if
Incidentally, when we
A Matter of Record (301) 890-4188
51
1
they're able to achieve higher exposures.
2
We pooled the data from these three phase 2
3
studies and tried to see if this dose titration had
4
been implemented based on individual patient
5
tolerability.
6
around how the titration was specifically
7
implemented.
8
dose titration had been implemented in the phase 2
9
studies, if it indeed translated to the right group
10 11
And I'll go through the details
But we wanted to see if the way the
of patients getting dose titrated. What we have here is the typical range of
12
exposures that you get at the 5-milligram BID dose
13
when a group of patients receive that particular
14
study dose.
15
large intersubject variability.
16
And you can see there is a fairly
From our DCE-MRI study, we arbitrarily
17
picked 150 nanograms hour per mL, which was half of
18
that 300 cutoff, which was for AUC 24.
19
picked that as our threshold for activity.
20
can see while the majority of patients are above
21
that, a subset of patients are getting
22
subtherapeutic exposures.
Then based on
A Matter of Record (301) 890-4188
So we And you
52
1
tolerability, we took all of these patients in each
2
of these studies.
3
systematically go up from 5 to 7 milligrams BID,
4
and then to a maximum of 10 milligrams BID if they
5
were able to tolerate the drug.
6
They were allowed to
As I mentioned, before titration, the
7
results were fairly variable.
Now, after we had
8
implemented titration, we landed up with three
9
groups of patients.
One was the group of patients
10
who did not need dose titration presumably because
11
they already had appropriate drug levels.
12
drug related effects fairly early, and so they were
13
not able to tolerate higher drug levels.
14
They saw
If you look at that group of patients which
15
did not need dose titration or were not eligible
16
for dose titration, you can see, for the most part,
17
again, the majority of them are above this
18
particular threshold.
19
at the bottom here who get lower concentrations.
20
There are still a few people
But the more interesting piece is that if
21
you look at -- and Dr. Mehrotra went over this
22
yesterday as well -- the 5-milligram BID before
A Matter of Record (301) 890-4188
53
1
titration exposures in the patients who eventually
2
went to either the highest 7-milligram dose or the
3
10-milligram dose, you see that those patients
4
started off with much lower exposures.
5
after titration, after they were allowed to go up
6
to the 7 milligrams BID or 10 milligrams BID, they
7
were able to get exposures similar to everybody
8
else.
9
And then
I like the picture, but if you'd like to see
10
the numbers here, it makes exactly the same case.
11
Those people who don't need dose titration, those
12
other exposures, and after titration at the 7- and
13
the 10-milligram BID dose groups, they're able to
14
get some more exposures.
15
One thing that we were very clear in
16
communicating in our package to regulators was that
17
when you do dose titration, when you go from 5- to
18
10-milligrams BID, you do not get a doubling in
19
exposure.
20
patients who had lower than typical exposures to
21
begin with, and you're allowing them to catch up
22
with the remaining patients.
All you're doing is you're taking those
A Matter of Record (301) 890-4188
54
1
This pharmacokinetic data confirmed that
2
what we are doing is normalizing plasma exposures
3
with dose titration.
4
Another analysis that we did was that,
5
again, we used this arbitrary threshold of
6
150 nanograms hour per mL for the AUC based on our
7
DCE-MRI results.
8
look at the patients who get higher than this
9
threshold and those who get lower than that.
And we said, well, okay, let's
10
If we pool the data from patients who had
11
higher than this threshold level and compare them
12
to those who had lower, you see a pretty good
13
separation in terms of their outcome.
14
fairly decent change in PFS going from 32 weeks to
15
52 weeks with a hazard ratio of .55, again
16
indicating that probably the right group of
17
patients stand to benefit from it.
18
And again, a
So what were the criteria for dose
19
titration?
How did we actually specifically
20
implement this?
21
patients needed to meet, and they needed to meet
22
all three of them in order to be able to dose
There were three criteria that
A Matter of Record (301) 890-4188
55
1
titrate.
2
the patient should not have had any drug related
3
adverse events, which were CTC grade 2 or higher
4
during at least 2 weeks of treatment, with a 5-
5
milligram BID starting dose.
6
rational.
7
The first one was fairly obvious, that
That's fairly
The second requirement was that we wanted
8
the patients to be normotensive.
This class of
9
agents causes increase in blood pressure.
It's a
10
very well known effect for anti-angiogenic VEGF
11
receptor-2 drugs.
12
an increase in blood pressure, then probably they
13
don't stand to benefit from further increases in
14
their dose.
15
And if a patient has already had
The third piece was that the patients should
16
not have been taking any anti-hypertensive
17
medication, again, with the understanding that
18
they've probably already got the increase in blood
19
pressure, and they've started taking
20
anti-hypertensive medication.
21 22
The other reason why we included that third criterion was that we recognized that there are
A Matter of Record (301) 890-4188
56
1
actually a lot of people who came on the study with
2
preexisting hypertension.
3
come on the study with preexisting hypertension
4
that is controlled with antihypertensive drugs.
5
And at that point, when we implemented dose
6
titration, we did not know if those group of people
7
were more sensitive to having an increase in blood
8
pressure when they were put on an anti-angiogenic
9
drug.
10
So there are people who
We now have data for more than a thousand
11
patients, where we have compared the increase in
12
blood pressure seen in those patients who come on
13
with a preexisting, controlled, hypertension versus
14
those who don't come on with preexisting
15
hypertension.
16
effect is almost identical in both of those groups.
17
So had we to implement it right now, we would
18
probably not included that third criteria.
19
that point, we wanted to err on patient safety, and
20
we included that as the third requirement.
21 22
And we actually see the drug related
But at
Of course, in all of our protocols, we were very clear that clinical judgment ultimately
A Matter of Record (301) 890-4188
57
1
overrode all of these requirements, which was
2
specified in the protocol.
3
but each individual investigator had to decide if
4
their patients stood to benefit from increasing the
5
drugs.
6
These were guidances,
How does this play out in the real world, if
7
you will?
This was our phase 3 study, which was
8
done against sorafenib.
9
recruited to the active arm with axitinib.
About 359 patients were And
10
what you can see is that out of 100 percent of
11
patients who received the 5-milligram BID starting
12
dose, about 28 percent of them remained at that
13
starting dose, presumably because they had some
14
drug related effect, which prevented them from dose
15
titrating.
16
About 34 percent of the patients actually
17
dose reduced below the 5-milligram BID starting
18
dose, and we had a lot of discussion yesterday
19
afternoon on whether 33 percent is the right number
20
for deciding MTD; it seems arbitrary.
21
as it may, if everybody is traditionally using that
22
33 percent, in this large study we were able to
A Matter of Record (301) 890-4188
But be that
58
1
confirm that 5-milligram BID is indeed the MTD
2
because almost exactly a third of the patients need
3
to dose reduce.
4
interest, which is 38 percent of the remaining
5
patients who did dose increase.
6
about 40 percent of the patients benefit from dose
7
titration.
8 9
Then this is the other group of
So on average,
One question that we got fairly early on from people, both internally as well as externally,
10
when we interacted with regulators was, well, your
11
dose titration is based on individual patient
12
tolerability.
13
monitoring?
14
done it.
15
but I think it stands to reason that maybe a
16
concentration driven clinical study might provide
17
better response than the one that we implemented in
18
our clinical development plan.
19
Why not just do therapeutic drug Of course.
Why not do it?
We haven't
We haven't systematically evaluated it,
There are two other things which make
20
therapeutic drug monitoring slightly challenging
21
for axitinib.
22
short plasma half-life, so it does not accumulate
One is that axitinib has a fairly
A Matter of Record (301) 890-4188
59
1
substantially with continuous dosing.
2
particular dosing interval, the concentrations go
3
up and come down fairly quickly.
4
come down.
5
you can collect a single blood sample and use that
6
as a marker of that particular patient's exposure.
7
And during a
So they go up and
So there is no one time point at which
So logistically in terms of implementing it
8
in the real world, you might have to collect more
9
than a few samples, which might present a
10
challenge.
11
your standard one blood sample drawn to drive your
12
dosing decision.
13
Not insurmountable, but it wouldn't be
Then the second piece of course is that we
14
don't have the specific plasma concentration
15
targeted for therapeutic monitoring.
16
of the analysis that we've done more recently with
17
the prospective study that I'm going to talk about
18
in just a few minutes, that particular analysis
19
tells us that, actually, there's no such thing as a
20
target concentration for this drug
21 22
In fact, some
But actually the target concentration varies from patient to patient, that different patients
A Matter of Record (301) 890-4188
60
1
might actually need a different target
2
concentration depending on either their VEGFR or
3
the vasculature of the tumor.
4
little naive for us to think that PK alone is
5
what's driving clinical response.
6
some impediments to implementing dose titration.
And it might be a
So there are
7
The other piece that we got a lot of
8
questions on, when do you implement dose titration?
9
We don't know when.
This is how we did it.
We
10
believe it works.
11
doing it.
12
scheme for dose titration based on what we knew at
13
that time.
14
patients who get higher exposures do better.
15
we knew that, at least in the clinical studies, the
16
first on-study scan happens at week 6 or week of
17
treatment.
18
going to come off the study.
19
There might be another way of
But I can tell you that we picked our
And what we knew at that time was that And
And once somebody progresses, they're
So we wanted to give patients an adequate
20
time to achieve this optimal exposure, and
21
therefore we wanted to implement the dose titration
22
fairly early.
So we said the earliest you can
A Matter of Record (301) 890-4188
61
1
implement dose titration is after 2 weeks of
2
dosing.
3
could happen at much time, at anytime later.
4
fact, a lot of the investigators prefer to wait and
5
see how their patients did before they implemented
6
dose titration.
7
do it was at 2 weeks.
It In
But the earliest they were able to
Also, like I said, axitinib has a short
8 9
It doesn't have to be at 2 weeks.
half-life.
It reaches steady state very quickly,
10
so we don't need to wait for a long time.
And most
11
of the adverse events for this class of agents
12
emerge fairly rapidly.
13
predictable effects, which emerge in the first
14
cycle of treatment.
15
we implemented it fairly quickly.
They emerge their reliable
So again, several reasons why
Every time we presented dose titration as
16 17
what had been implemented in our study, our team
18
was surprised people were relatively accepting of
19
it.
20
your retrospective analysis kind of makes sense.
21
But we also got some criticisms in terms of people
22
saying, well, after all, those are retrospective
It made rational sense.
People said, yeah,
A Matter of Record (301) 890-4188
62
1
analyses.
People who stay on the study for a
2
longer period of time are more likely to have drug
3
related effects, and therefore are more likely to
4
dose titrate.
5
analysis may be biased in favor of those people who
6
stay on the study for a longer period of time.
So inherently, your retrospective
7
So in order to convince ourselves, indeed,
8
that this was the right thing to do, while we were
9
putting together the package for this particular
10
drug, we implemented a prospective study to once
11
and for all evaluate the benefit of dose titration.
12
And this was a randomized, double-blind study with
13
200 patients in first-line RCC.
14
This was the design of the study.
All 200
15
patients initially received the 5-milligram BID
16
starting dose.
17
that we had in all of our studies, including our
18
phase 3 study, we allowed a 4-week period for
19
patients to kind of stabilize on their 5-milligram
20
BID dose.
21
eligible for titration using very similar criteria.
22
If they were not eligible for titration, then they
And instead of the 2-week period
And then we evaluated them if they were
A Matter of Record (301) 890-4188
63
1
were allowed to continue at the 5-milligram BID
2
dose.
3
drug related effects during the first month of
4
treatment.
These were people who presumably had some
If they were eligible for titration, then
5 6
they were randomized into one of two groups.
7
both groups, they received 5-milligrams BID
8
axitinib group, but in one group, they additionally
9
dose titrated with active drug, and in the other
10
In
group they titrated with active placebo. Now, the primary endpoint of this study was
11 12
ORR.
13
primary endpoint was PFS, but because only a subset
14
of patients are eligible for titration and they'd
15
further break down into two groups, we would have
16
had to start with 800-, 900-patient study, and we
17
obviously did not have the stomach for doing a
18
900-patient study for this particular purpose.
19
We would have loved to do a study where the
So our primary endpoint was overall
20
survival -- sorry, was ORR, and PFS, OS, safety,
21
and duration of response were secondary endpoints.
22
We see response rates in the upper 50 percent
A Matter of Record (301) 890-4188
64
1
range, so with the 200 patients and expecting only
2
a subset of them to dose titrate, this still had
3
80 percent power to detect at least a 25 percent
4
increase in the response rate.
5
We also did a very thorough evaluation of
6
axitinib PK as well as blood pressure.
We did
7
ambulatory blood pressure monitoring over 24 hours
8
to really characterize the right time for measuring
9
blood pressure, et cetera.
So this was our
10
biomarker heavy study where we also included a lot
11
of pharmacogenomic assessments.
12
In terms of the patient characteristics,
13
fairly standard.
14
male, 61 median years of age.
15
Caucasian.
16
because Japan participated in this study as well.
17
The majority of the patients were Most of them were
And then we had some Asian subjects
In terms of the results, what did the
18
pharmacokinetic data tell us?
19
exposures at the 5-milligram BID starting dose,
20
before they titrated, and divided the patients into
21
two groups, those who were eligible for titration
22
and those who were not eligible for titration,
A Matter of Record (301) 890-4188
If we look at the
65
1
fairly dramatic difference, just as we expected,
2
that the patients who were eligible for titration
3
had much lower exposures to begin with.
4
therefore, post-titration, they stood to benefit
5
and normalize to what the other group got.
6
these are the same results in the figure.
And
And
7
As I mentioned earlier, blood pressure is a
8
very reliable, common -- increase in blood pressure
9
is a very liable, common effect seen with this
10
class of drugs.
11
evaluated blood pressure very rigorously in this
12
study, and wanted to look and see how the efficacy
13
compared in people who had the increase in blood
14
pressure versus those who did not.
15
Again, we very systematically
We had three separate measures for it.
We
16
looked at the change in blood pressure, either a
17
10-millimeter increase in mercury or a
18
15-millimeter increase in mercury in diastolic
19
blood pressure, or an increase in the absolute
20
value above 90.
21
see that the group which had an increase in blood
22
pressure did better, both in terms of their ORR as
And in each of the groups, you can
A Matter of Record (301) 890-4188
66
1
well as median PFS. Even better if you look at the PKs within
2 3
the subgroup of patients.
4
increase in blood pressure in general had higher
5
exposures than the others.
6
with exposure, PK, and blood pressure, fit in very
7
nicely.
8 9
Individuals who had the
So really, the piece
So this was the proof of the pudding.
This
was the primary endpoint of the study, which was
10
overall response rate.
11
the groups in the study, the overall response rate
12
was about 48 percent, and these two are the
13
predominant comparators.
14
which was actively titrated, 54 percent response
15
rate, the group which was titrated with placebo,
16
34 percent.
17
If we pooled data from all
If you look at the group
This was statistically significant.
The
18
study met its primary endpoint.
19
you look at the group which did not need to be
20
titrated, presumably, because they had the
21
appropriate exposures to begin with at the
22
5-milligram BID group, they did the best; not
A Matter of Record (301) 890-4188
Interestingly, if
67
1 2
unsurprisingly. If you look at the progression-free
3
survival, we had only 50 patients in each arm.
4
the active and placebo titration group, we really
5
did not see a difference.
6
you, in our phase 3 study, we had almost 400
7
patients to look at a difference of about 2 to 3
8
months in terms of progression-free survival.
9
Now, again, to remind
So with a substantially smaller number of
10
patients, we were not able to see a difference
11
between those three groups.
12
though, even though this difference was not
13
statistically significant, the hazard ratio was
14
0.85 in favor of dose titration.
15
In
I should point out,
I should also mention that the survival data
16
from this study has just matured.
17
couple of weeks at ASCO, we will be showing results
18
from the survival endpoint in Chicago.
19
In fact, in a
In terms of the safety results, these were
20
the safety results.
This is the total group, the
21
group which received active titration, the group
22
which received placebo titration, and then the
A Matter of Record (301) 890-4188
68
1
non-randomized group. There are only a few things I want to point
2 3
out.
4
pretty comparable across three of these groups.
5
The only place where we saw really substantial
6
differences was for hypertension as well as
7
hand-foot syndrome, both of which, again, are
8
classic adverse events you see with this class of
9
agents.
10
If you look at the AEs, they were actually
Therefore, it stands to reason that the
11
group which got active titration probably had
12
higher exposures, and therefore saw more
13
hypertension, and likewise, more PPE,
14
palmar-plantar erythrodysesthesia.
15
fact you see that the group which was not
16
randomized actually had the highest level of
17
increase in blood pressure.
18
And then, in
So what were the conclusions from our
19
prospective study?
What we concluded was that
20
patients who titrated with active axitinib
21
experienced a significant improvement in overall
22
response rate, 54 percent versus 34 percent.
A Matter of Record (301) 890-4188
We
69
1
could not really make any conclusions for the PFS,
2
although we do know that the study was not powered
3
for PFS, the hazard ratio did favor dose titration.
4
We also know that patients who were titrated,
5
eventually started over with much lower exposures,
6
just as we had predicted.
7
I guess what the study convinced us is that
8
the particular dosing algorithm that we used all
9
through the clinical development of the program,
10
what was used in the phase 3 study, what is
11
currently on the label, is able to identify that
12
subgroup of patients who stand to benefit from
13
increase in dose titration.
14
Is this the perfect scheme?
Probably not.
15
Do we stand to benefit by fine tuning this dose
16
titration scheme?
17
drug monitoring is the answer.
18
need to do more work to figure it out.
19
is with this scheme at which it was developed and
20
as it is on the label, it does seem to work.
21
with that, I want to end with acknowledgments, and
22
thank you for your attention.
Absolutely.
A Matter of Record (301) 890-4188
Maybe therapeutic We don't know.
We
All we know
And
70
1
(Applause.)
2
DR. LIU:
3
Thank you, Yazdi.
That was a
really good presentation. We're right on time, and now we're going to
4 5
have a 30-minute break, and we will restart at
6
9:45.
(Whereupon, at 9:18 a.m., a recess was
7 8 9
Thank you.
taken.) DR. BULLOCK:
We're going to go ahead and
10
start the second half of our morning session.
11
kick off, we're going to have Eric Masson, who's
12
the global lead for quantitative clinical
13
pharmacology at AstraZeneca, tell the story of
14
vandetanib.
15 16
To
Presentation - Eric Masson DR. MASSON:
Good morning, everyone.
I'm
17
glad everyone is back and happy to be here to
18
present a case study on vandetanib.
19
you today is that sometimes for some
20
compounds -- in contrast to what we here, sometimes
21
the appropriate dose is the MTD.
22
start, let's talk about the drug development
A Matter of Record (301) 890-4188
What I'll show
But before we
71
1 2
concept and how it applies to oncology. The learning confirm approach, as defined by
3
Lew Sheiner, in the drug development, we learn and
4
confirm.
5
molecule, its property, pharmacokinetic property,
6
the safety of the compound.
7
target engagement.
8
that are biologically active to confirm efficacy in
9
phase 2.
10
So in phase 1, we learn about the
We need to explore
And from that, we select doses
At each stage of development, as is shown on
11
the lower curve, based on the profile, we put gold
12
posts, do we achieve what we were expecting, either
13
from a safety perspective, a PK, or efficacy.
14
we see a profile that looks like the green curve,
15
where there's clearly differentiation, then we move
16
forward to the next step.
17
If
Sometimes we hit a red signal where the
18
compound clearly is not competitive, and then we
19
stop development.
20
kind of in between where we still don't know and we
21
still further explore.
22
that's followed by phase 2B, where we learn about
And then sometimes often we're
For most therapeutic areas,
A Matter of Record (301) 890-4188
72
1
the dose response, and then phase 3, where we
2
confirm the efficacy and safety and the appropriate
3
dose.
4
What's happening in oncology because of the
5
unmet medical need is that cycle gets truncated
6
many times.
7
diseases in oncology, based on activity, many
8
compounds get filed based on phase 2.
9
means that you need to characterize the dose and
So what you see is that for certain
And that
10
exposure response early on in your phase 1 study.
11
So I'll describe vandetanib today as a case study.
12
Another concept is the therapeutic index or
13
the clinical utility index.
14
your development, you explore different doses, and
15
you characterize the efficacy on the left part.
16
And with more data, you reduce the uncertainty.
17
Yesterday, we heard a speaker -- Dinesh mentioned
18
that actually by using a continuous endpoint, you
19
can actually reduce variability as opposed to using
20
T tests or categorical endpoints.
21 22
Hopefully, throughout
Likewise, for safety, by having more data, you would use the uncertainty.
A Matter of Record (301) 890-4188
And then you can
73
1
use that data to build up a clinical utility curve
2
to look at the balance between efficacy and safety
3
versus dose and exposure to characterize your
4
therapeutic window.
5
typically more emphasis on the efficacy, although
6
safety is important as well.
7
For oncology, there's
What are some therapeutic concepts to
8
improve the therapeutic index for oncology
9
compounds?
I think it's been clear that we need to
10
do more dose and exposure assessment early, looking
11
at efficacy and safety, and ideally looking at
12
longitudinal endpoint because many of our endpoints
13
are categorical.
14
categories.
15
shrinkage, for example, you can get better insight
16
into the dose and exposure relationship.
17
They get transformed into
But by looking at the whole tumor
We need to explore schedules.
For some
18
targets, you need a continuous hit to block the
19
pathway.
20
targets, like PI3 kinase, MEK, and others, you can
21
give intermittent dosing.
22
built up from proper preclinical experiments to
But we know that for several other
And that needs to be
A Matter of Record (301) 890-4188
74
1
look at which schedule can still give you efficacy
2
and actually can reduce your toxicity profile.
3
Sequence versus concomitant might be
4
important when you start combining compounds.
5
That's something you can also learn from
6
preclinical.
7
Food effect should be conducted early in
8
oncology, and sometimes it's difficult because you
9
cannot do healthy volunteers, so you have to build
10
up these questions within, let's say, a dose
11
expansion cohort to not only determine the impact
12
of food on exposure, but also the variability, as
13
well as the tolerability.
14
improve GI tolerability.
15
can build that in your phase 2 program.
16
Sometimes food will By doing that early, you
Sometimes you need to look at
17
pharmacogenetic, depending on the pathways, to
18
characterize metabolic pathways or you can also
19
profile the tumors to see which patients are more
20
likely to benefit.
21
heard a nice case study from Yazdi, and that's
22
possible when you have a clear PD biomarker and
Dose titration is possible.
A Matter of Record (301) 890-4188
We
75
1 2
typically a short half-life compound. Finally, I think we've improved tolerability
3
of many of our drugs because of the treatment,
4
concomitant meds that allow us to better improve
5
therapeutic index.
6
drugs that can help that.
7
For GI or neutropenia, we have
Today I'll talk about vandetanib.
8
Vandetanib is a multikinase inhibitor.
It hits
9
several pathways important for cancer, including
10
VEGFR-2, EGRF, and RET.
11
was clearly shown that the efficacy is dose and
12
exposure dependent, so you need to give maximum
13
dose to get tumor regression.
14
In preclinical studies, it
What we learned from our phase 1 program was
15
that the dose was set at 300 milligrams once daily.
16
The drug is slowly absorbed with a Tmax of 6 hours
17
and there was no food effect in our food effect
18
studies conducted in healthy volunteers.
19
What's unique about this drug is this drug
20
has a very large volume of distribution and has a
21
very long half-life for small molecules.
22
half-life is almost 3 weeks 19 days.
A Matter of Record (301) 890-4188
So its
As a result,
76
1
we give the drug daily, but there's an 8- to
2
10-fold accumulation, and we reach steady state in
3
about 2 months of treatment.
4
In our early program, the DLTs that we saw
5
were GI, diarrhea, skin rash, and hypertension.
6
But we also detected QT signal because we had
7
intense ECGs on our phase 1, and we warned further
8
that there was a clear QTc prolongation that was
9
dose and exposure dependent, and I'll show you some
10 11
of that data. The drug was approved in 2011 by the FDA and
12
in 2012 by EMEA, based on a phase 3 study that
13
shows improved PFS in patients with advanced
14
medullary thyroid cancer or MTC.
15
approved for the treatment of symptomatic or
16
progressive medullary thyroid cancer in patients
17
with unresectable, locally advanced or metastatic
18
disease.
19
The drug is
The use should be in patients with indolent,
20
asymptomatic, or slowly progressing disease, and
21
should be done after consideration of treatment
22
risk of the drug.
The dose that's approved is
A Matter of Record (301) 890-4188
77
1
300 milligrams daily.
2
except for patients with moderate or severe renal
3
impairment, which needs to be started and reduce
4
those of 200 milligrams.
5
it's clearly mentioned that both -- so you start at
6
300, but based on tolerability, you can dose reduce
7
for the management of side effects.
8 9
That's for most patients
However, in the label,
This is the development program for thyroid cancer of vandetanib.
There were two phase 1
10
studies conducted, one in healthy volunteers where
11
we dosed from 300 to 1200 milligrams, 23 healthy
12
volunteers.
13
dose in patients where the MTD was established at
14
300 milligrams.
15
had a minimum of 3 patients at each dose.
16
we hit toxicity, then we expanded to get more
17
confidence about the safety profile.
18
cohort of 100 milligrams and 300 milligrams that
19
were further expanded to learn more about the
20
safety of the drug.
21 22
There was an MAD multiple ascending
This was not a pure 3 plus 3.
And once
There was a
Based on that data, because of the inhibition of RET, which might be important for
A Matter of Record (301) 890-4188
We
78
1
thyroid cancer, there's an investigator sponsor
2
trial that was conducted using 300 milligrams in
3
patients with advanced medullary -- MTC.
4
study was done in 30 patients, and it shows
5
efficacy with an overall response rate of
6
20 percent.
This
We did then do a subsequent study at a lower
7 8
dose of 100 milligrams in 19 patients, which also
9
showed you some efficacy but a slightly lower
10
response rate of 15 percent.
11
confirmatory study, phase 3, that was conducted
12
where patients were randomized in a 2 to 1 ratio to
13
receive, in a double-blind fashion, either
14
vandetanib or placebo in patients with advanced
15
MTC.
16
were 231 patients treated on vandetanib and 100 on
17
placebo.
18
because this is a rare disease.
19
Then there was a
That trial is called the ZETA trial.
There
That study took four years to complete
At the time of approval, the FDA requested
20
to do a postmarketing study.
It was a phase 2
21
study where patients were randomized 1 to 1 to
22
receive, in a double-blind fashion, either
A Matter of Record (301) 890-4188
79
1
150 milligrams or 300 milligrams of vandetanib.
2
There were 81 patients in that study, and I'll show
3
you the preliminary results of that study today.
4
On this slide, you see the results that led
5
to the approval of the drug by the FDA and EMEA,
6
and this is the PFS showing that vandetanib, on the
7
yellow curve, clearly improved progression-free
8
survival over placebo, with a hazard ratio of 0.46
9
and a p-value that was highly significant.
10 11
This
study was published in JCO in 2011. Using data from this study, we did exposure
12
versus QTc analysis.
So there were 231 patients
13
who received vandetanib in that study.
14
panel, you see the plasma concentration profile
15
over time, so you see that the concentration
16
steadily increased in the first 60 days of
17
treatment and then reached steady state to be
18
maintained further on.
19
the QTc signal with time, and you see that the QTc
20
signal, the QTcF increased from baseline in the
21
first 6 weeks of treatment, and then reached a
22
plateau.
On the left
On the right panel, you see
A Matter of Record (301) 890-4188
80
1
We did exposure versus QTc analysis, so we
2
fitted this data using an Emax model, and showed
3
that -- you can see the curve on the right with the
4
predicted interval.
5
with concentration and reached a plateau at around
6
750 nanogram per mL, and that the mean signal at
7
the 300 milligrams was actually 34, merely second
8
increase.
9
actually there's a box warning in the label.
10
You see that the QTc increased
So clearly, a clear QTc signal.
And
Now, we also did exposure versus efficacy
11
from that study using three categories, partial
12
response, shown on the blue curve.
13
that the probability of partial response increased
14
with increasing exposure, whereas on the red curve,
15
the progressive disease rate decreased with
16
increasing exposure.
17
represents the average exposure at steady state
18
with 300 milligrams daily.
You can see
And the vertical bar
19
What did we learn from these analyses?
20
there's a clear QTc signal that's depending on
21
concentration.
22
signals, but we could not detect -- probably
We also looked at other safety
A Matter of Record (301) 890-4188
That
81
1
because of the low rate, we could not detect a
2
clear relationship with other safety events.
3
only QTc was there a clear exposure response.
4
Likewise, for efficacy, we could detect a clear
5
exposure response for efficacy using overall
6
response versus progressive disease, but we could
7
not detect relationship between progression-free
8
survival or overall survival.
9
So
Study 97 is the postmarketing commitment
10
study that was conducted.
11
conduct this study, and you can look it up in the
12
summary basis of approval.
13
a trial to explore alternative vandetanib dose or
14
dosage regimen that will reduce toxicity but
15
maintain the efficacy compared to 300 milligrams.
16
This was conducted in a randomized fashion in the
17
same populations with advance MTC, and we compared
18
2 doses, 300 and 150 milligrams.
19
The FDA requested to
The goal was to conduct
The primary endpoint is overall survival.
20
The safety should include -- and this was the FDA's
21
request -- evaluation of vortex keratopathy and
22
corneal stromal change with ophthalmologic
A Matter of Record (301) 890-4188
82
1
examination every six months.
And the safety
2
should also evaluate heart failure using a serial
3
echocardiogram in all patients. This is the title of the study that was
4 5
conducted.
And bear in mind, this study was not
6
designed a formal comparison.
7
small study.
8
you probably will never hear about it because it
9
will take forever to enroll or reach, let's say,
This is a relatively
And if we have to power the study,
10
progression-free survival or overall survival
11
endpoint. This is the design.
12
Patients were
13
randomized 1 to 1 to receive either vandetanib 150
14
or 300.
15
starting dose of 150 and 41 patients to receive a
16
starting dose of 300 milligrams.
17
select the cutoff for the interim analysis at 14
18
months, and that's because some patients take
19
several months to respond, up to a year, and
20
sometimes even beyond.
21
majority of the responses will be achieved within
22
14 months.
There were 40 patients who received a
We chose to
And we thought that the
So that's why the primary data cut was
A Matter of Record (301) 890-4188
83
1 2
done at 14 months. After 14 months, patients had the option of
3
either continuing at the dose they tolerated,
4
whether it was 150, 300, or a reduced dose.
5
then patients who had a response at 150 had the
6
options of also escalating to 300.
7
part B, which is still ongoing, we're only
8
collecting safety.
9
And
However, for
So today, I'm going to show you the interim
10
results from part A with a data cut at 14 months.
11
The primary objective is to look at overall
12
response, and then the secondary objective is to
13
look at safety tolerability, also looking at other
14
endpoints such as objective response, and also
15
looking at maximum tumor shrinkage.
16
looking at the PK also and QTc.
17
And finally,
I'm going to first show the safety results
18
from that study.
Almost all patients reported
19
adverse events, and there were a similar proportion
20
of patients reporting serious adverse events at
21
either dose.
22
thought not to be related to the drug by the
There were 2 deaths, but it was
A Matter of Record (301) 890-4188
84
1
investigator in the 300-milligram group.
Dose
2
reduction was required in 15 percent of the
3
patients at 150 and in 29 percent of patients at
4
300.
5
percent tolerated the initial dose of 150 as
6
opposed to 71 percent at the 300 milligrams.
That means the majority tolerated, so 85
The adverse events that led to
7 8
discontinuation were higher in the 300-milligram
9
cohort, at 12 percent versus 5 percent.
And the
10
most common adverse events were similar to the
11
previous experience we had.
12
adverse event.
13
about the compound.
14
or higher adverse event at 300, but when we look at
15
drug related grade 3 and above, they were similar
16
between the two groups.
So there was not a new
It was consistent with what we knew More patients reported grade 3
When we look at causally related adverse
17 18
events or drug related adverse events, the most
19
commonly observed drug related adverse events were
20
generally higher in the 300-milligram cohort
21
compared to the 150-milligram as shown here on this
22
table.
There were more diarrhea, rash,
A Matter of Record (301) 890-4188
85
1
hypertension, keratopathy, as well as QTc
2
prolongation. There was a more increase in blood pressure.
3 4
This is also a VEGF inhibitor, and so there was
5
more observed at the 300-milligram group, and
6
there's also more QTc prolongation, which occurred
7
sooner at the 300-milligram group.
8
were no occurrence of Torsades de Pointes, and on
9
the ophthalmology exam, there was a slow increase
However, there
10
in condition over time, with slightly higher
11
prevalence at the 300-milligram cohort.
12
there were no findings of cardiac failure with
13
ejection fraction below 40 percent observed in this
14
study.
15
Finally,
Now, let's switch to the efficacy.
The
16
overall response rate was 20 percent with the
17
150 milligram and 29 percent for the 300 milligram.
18
Most of them were partial response.
19
of the patients had stable disease, so 52.5 percent
20
at 150 versus 56 percent at 300 milligram.
21
we can see, there were more progressive disease
22
observed at the 150 milligram with 22.5 percent
A Matter of Record (301) 890-4188
The majority
And as
86
1
progressing versus 5 percent at the 300 milligram.
2
Again, this was not powered to show a difference
3
but to observe the efficacy and safety at two
4
different doses.
5
When we look at the maximum change in tumor
6
lesions, the waterfall plots are shown here.
7
top shows the 150 milligram and the bottom the
8
milligram.
9
there's actually more tumor shrinkage observed
10 11
The
So you can see at 300 milligram,
compared to the 150 milligram. Finally, the overall conclusion is that the
12
current dose regimen for Caprelsa, as described in
13
the label, allows for dose reduction in patients
14
who don't tolerate the initial starting dose of
15
300 milligrams.
16
This study still supports the dose of 300
17
once daily, allowing dose reduction in patients who
18
don't tolerate the starting dose.
19
150 milligram, study 97 evaluated a lower dose, but
20
we believe it did not change any conclusion of our
21
current regimen as described in the label.
22
In the
Before I leave, I'd like to thank many
A Matter of Record (301) 890-4188
87
1
people who have contributed to this work,
2
especially the team, Kevin Heller, who is actually
3
here today, medical director; Karen Mitz, Paul
4
Martin, and Martin Johnson, in our function
5
supporting clinical pharmacology and
6
pharmacometrics; as well as the senior management
7
at AZ for allowing me to present this data, my
8
boss, Don Stanski, as well as other leaders in
9
oncology. I'd like to thank especially the
10 11
investigators and their staff for conducting these
12
studies.
13
conducted, and especially thanks to the patients
14
and their families who have accepted to participate
15
in these trials.
16
AACR and FDA organizing committee for this
17
invitation to share this data.
These are not easy studies to be
Finally, I'd like to thank the
18
(Applause.)
19
DR. BULLOCK:
Thank you.
Thank you, Eric.
That was
20
great.
I would like to introduce Dr. Jin Jin,
21
who's the associate director and head of modeling
22
and simulation at Genentech, who's going to talk
A Matter of Record (301) 890-4188
88
1
about some of the options for combination therapies
2
and dose optimization. Presentation - Jin Jin
3 4
DR. JIN:
Thank you, Julie.
5
Good morning, everyone.
Firstly, I would
6
like to start as thanking Judy and Qi for inviting
7
me to be part of such a great workshop and also
8
thank you to all the fellow speakers and panelists.
9
I have truly really enjoyed the learnings and all
10
the very dynamic discussions, and really looking
11
for more. My topic today is actually the dose
12 13
optimization for small molecule combinations in
14
oncology.
15
oncology going to polypharmacy, and today I'm going
16
to share some of the collective learnings that we
17
have had in Genentech in the past few years.
This is really a very growing area in
So I'll start with a very brief overview,
18 19
talking about some of the unique challenges and
20
opportunities in combination therapy and also what
21
are the dose optimization strategies in this unique
22
area.
And then, rather than the fellow speakers in
A Matter of Record (301) 890-4188
89
1
the session, which everyone shared exciting stories
2
from one project, today I'm going to share
3
collective learnings from several projects in this
4
area for doing dose optimization for combinations
5
from different aspects for PK, efficacy, and safety
6
aspects, as I shared on the diagram on the lower-
7
right corner. From the PK perspective, addressing
8 9
combination and also potential drug-drug
10
interactions, using a physiological based PK
11
approach and also population PK approach; for
12
efficacy, using translational PK/PD, clinical
13
PK/PD; and also a biomarker approach for optimizing
14
combination dose; and at the very end, last but not
15
least, for tolerability, which is even more
16
critical, especially for small molecule
17
combinations, how we can specially benchmark
18
leveraging historical information from single
19
agent.
20
some future directions hopefully to trigger more
21
discussion at the panel discussion.
22
At the very end, a very brief summary and
Why do we need to do combination therapy
A Matter of Record (301) 890-4188
90
1
oncology?
I think this is probably -- I don't need
2
to even give an answer for this audience.
3
really, as you can see, there are clearly, for the
4
kinase pathways on the left side of the slide, it's
5
very much heavily targeted pathways.
6
stars highlight the target that has been either
7
investigated or approved as drug targets, and I'm
8
sure there are several targets that I've missed on
9
the diagram.
But
The yellow
The vision of polypharmacy oncology is
10 11
really to have simultaneous inhibition of multiple
12
targets to enhance activity in broader populations
13
and also with less resistance.
14
scientific rationales are very clear.
15
multiple pathways downstream of validated targets.
16
There are prominent mutation types of several
17
different multiple mutations in different cancer
18
types.
Some of the There are
19
Also, there are multiple cross-talks and
20
feedback loops across these feedback mechanisms,
21
which sometime cause resistance when you are
22
treating the cancer as a single agent.
A Matter of Record (301) 890-4188
This calls
91
1
for combination therapy in oncology, and for
2
combination, it can be a combination of a new
3
molecular entity with standard of care or a
4
combination of two new molecular entities.
5
So what is the difference of these two types
6
of combinations?
First, the new molecular entity
7
standard of care combination, generally for the
8
standard of care, the dose and regimen are known.
9
We already have very rich collective knowledge
10
about the PK, efficacy, and safety of the molecule,
11
and we generally try to focus on optimizing dose
12
and regimen for the new molecular entity.
13
For the new molecular entity plus new
14
molecular entity type of combination, which is also
15
called 2NME combination, this is really a new
16
evolving area with accumulating knowledge.
17
this is combination of two molecules active in
18
clinical investigation and both have probably
19
limited knowledge regarding the clinical PK,
20
efficacy, and safety, it's really called a unique
21
case to optimize dose and regimen for both agents
22
simultaneously, at the same time.
A Matter of Record (301) 890-4188
Because
This actually
92
1
gives us an opportunity to do innovative clinical
2
development and also innovative quantitative
3
analysis to address such a type of question. Here, just listed out as a bullet, are some
4 5
of the challenges, specifically for 2NME
6
combination.
7
it's really, then, how to test two molecules.
8
of those do not have a rich clinical efficacy data
9
to tell you what is the right xenograft dose and
From a clinical advocacy perspective,
10
what is the right exposure target for both
11
molecules.
Both
Per clinical safety, as we have heard from
12 13
the session yesterday morning, translation for
14
toxicity from preclinical to clinical is already
15
challenging, and now for combination, it's even
16
more.
17
the field of do you need to do a combo on a safety
18
study?
19
not do it?
20
Also, there's a general question paradigm in
Is there value added?
Should we do it or
For clinical development, especially the
21
co-development of two molecules, the overall
22
development plan is complicated.
A Matter of Record (301) 890-4188
You need to
93
1
optimize the dosing strategy for both molecules.
2
The two molecules may have potential drug-drug
3
interaction from PK perspective, basically,
4
A affect B or B affect A.
5
perspective, that's where actually we're looking
6
for some drug-drug interaction.
7
the efficacy perspective, the two molecules work
8
better.
And also, from the PK
We're hoping from
There are some beneficial drug-drug
9 10
interaction from the efficacy perspective.
And
11
from the tolerability perspective, we are hoping
12
there is no manifesting type of tolerability when
13
you're giving two compounds at the same time.
14
is especially of critical important for small
15
molecule combination, where both of the compounds
16
are very likely to have a narrow therapeutic
17
window.
This
18
Also, with dosing conditions, I think we
19
have heard on both days about the importance for
20
dosing with food, but then now you have to two
21
molecules.
22
administered with food, but the other one is
What if one molecule is best to
A Matter of Record (301) 890-4188
94
1
without food?
2
molecules and co-administer at the same time?
3
are some of the smart ways to do it and also to
4
test that?
5
challenges; either it's co-formulation under
6
consideration or if you formulate separately, is
7
there any potential interaction between the
8
recipient?
9
How do you combine those two What
From the CMC area, of course some other
From the regulatory path perspective, this
10
is still a novel area for exploration.
11
regulatory path to explore the potential
12
registration for two new investigation drugs at the
13
same time?
14
for other therapeutic areas, which require
15
polypharmacy, such as for example, HIV, which is
16
very common for a cocktail type of treatment.
17
there has been several success examples of getting
18
approved drugs with co-formulation, for example.
19
But in oncology, these are new areas I think we are
20
all actively still exploring.
21 22
What is the
Some of these topics are also common
And
So how do we do dose optimization strategy oncology?
I think many of the previous speakers
A Matter of Record (301) 890-4188
95
1
have already talked about the topic, and what I'm
2
showing here is a schematic diagram linking the
3
dose exposure PK and also efficacy endpoint, and
4
these include biomarker, tumor response type of
5
early endpoint, and also late phase type of
6
endpoint, such as PFS and OS in oncology. Going vertically, from up to down, really
7 8
illustrates the translation from preclinical to
9
clinical, and from left to right is illustrating
10
the translation from early clinical to late
11
clinical.
12
challenges.
13
clinical translation, such as efficacy, you're
14
translating from generally homogenous xenograft
15
model to a very heterogeneous oncology patient
16
population.
Different translations oppose unique As we know from preclinical to
Also, the resistance development many times
17 18
is not very well captured in animal models.
19
safety, it is very challenging.
20
that.
21
are in broken dotted lines.
22
For
I'll just end with
That's part of the reason why the red arrows
From early to late clinical translation,
A Matter of Record (301) 890-4188
96
1
we're hoping to find predictive biomarkers.
I
2
think this is still a very fast-evolving area,
3
especially in oncology.
4
some cases, we are not that lucky.
5
translation from early tumor response to late
6
clinical outcome, this has been a topic touched
7
upon by several speakers in previous sessions as
8
well.
9
especially for the cases where you have very late
Some cases, we are lucky; Also,
Again, safety is very challenging,
10
onset type of adverse events or very chronic
11
adverse events.
12
findings from early clinical trials to late
13
clinical trials, this is still a rolling question
14
we are trying to answer.
How to translate the safety
As you can see, I tried to highlight the PK,
15 16
efficacy, and safety by different colors.
On the
17
upper-right corner, I have this diagram.
18
this diagram throughout my slides to illustrate the
19
different case examples targeting for different
20
areas.
21
strategy paradigm in oncology.
22
the complexity really will double or even triple
I use
This is a typical dose optimization
A Matter of Record (301) 890-4188
For combinations,
97
1 2
for combination therapies. For combination, we will start with PK for
3
especially assessing potential drug-drug
4
interactions for small molecule combinations.
What
5
are some of the potential reasons for PK DDI?
Many
6
times for small molecules, it's either metabolic
7
enzyme mediated, or transporter mediated, or
8
sometimes can be absorption mediated as well.
9
What are some of the different approaches we
10
may use to address this special challenge?
11
Physiologically-based PK, PBPK I think should be a
12
very familiar term for most of the audience here.
13
It's really a modeling and simulation tool by
14
integrating patient specific systems, specific
15
information, and also drug specific information
16
together, building into a quantitative modeling
17
framework, using this approach to try to predict
18
the effect of intrinsic and extrinsic factors.
19
This is scientific area with very fast
20
evolvement, especially in recent years, both from
21
scientific-based science, scientific evolvement,
22
from a wide replication in drug development by
A Matter of Record (301) 890-4188
98
1
multiple industries, and also in the regulatory
2
space, how to integrate this information as part of
3
the registration support and also labeling
4
language.
5
Shown here are just two diagrams pulling
6
from the literature, on the left-hand side,
7
illustrating the type of data this approach
8
integrates, and on the right-hand side, what type
9
of model structures and also the information this
10
approach can be used to address.
11
approach for combination therapy, we feel, as based
12
on internal experience, this approach can provide
13
very critical input on dose selection and
14
inclusion/exclusion criteria for clinical trials
15
for combination, especially for both the standard
16
of care plus new molecular entity combination or
17
2NME combination.
18
Using this
I'm going to show you two examples in the
19
coming slides, one for each case.
The first
20
example of application of PBPK is to inform on
21
paclitaxel combination trials for phase 1B, how to
22
inform the dose selection.
The challenge we have
A Matter of Record (301) 890-4188
99
1
here is we may have a potential drug-drug
2
interaction risk for paclitaxel, which is a
3
compound with no narrow therapeutic window.
4
opportunity here is whether we can prospectively
5
use PBPK approach or retrospectively use a
6
population PK approach, both to de-risk the DDI
7
perspectively and retrospectively confirm our
8
learnings.
9
The
Personally, I always feel although we have a
10
unique challenge, many times, the challenges can
11
easily translate to a unique opportunity for us to
12
implement some novel design and measurement
13
strategy together with smart analysis strategy to
14
give us a learning, because generally, these unique
15
challenges, the traditional approach doesn't work
16
as well.
17
In this case, we have a molecule of this
18
case, a PI3K inhibitor, which based on in vitro
19
study, there is an inhibition risk for CYP2C8.
20
paclitaxel is a known 2C8 substrate.
21
spite, based on the in vitro I/Ki assessment,
22
there's very low PK DDI risk.
A Matter of Record (301) 890-4188
And
Therefore, in
However, because of
100
1
the very narrow therapeutic window of paclitaxel,
2
the clinical impact may be high because it may
3
manifest the tolerability profile for paclitaxel.
4
So we want to get some refined assessment before
5
doing the phase 1B combo trials with PI3K and
6
paclitaxel.
7
Based on the PBPK simulation, rather than
8
paclitaxel, we used rosiglitazone as a surrogate
9
compound that is also a 2CA substrate.
The
10
analysis suggests that there is very low PK DDI
11
risk at the starting dose of the PI3K inhibitor in
12
the phase 1B study.
13
After getting the data from the phase 1B,
14
the readouts are also further confirmed by
15
retrospective analysis using population PK approach
16
to compare the observed paclitaxel PK in the combo
17
study comparing with historical PopPK prediction
18
from paclitaxel and vice versa, the observed PI3K
19
PK from the combo setting relative to the
20
historical single-agent PI3K population PK.
21
addition, we also didn't see any metabolite
22
providing changes for paclitaxel.
A Matter of Record (301) 890-4188
In
101
This is example, in this case, we used both
1 2
PBPK approach and the PopPK approach to address
3
this potential PK DDI risk for this combination.
4
The second example is a 2NME combination
5
case.
The challenge we have here is there are
6
potential bidirectional PK drug-drug interaction; A
7
affect B and also B may affect A.
8
really an iterative, dynamic process, and the
9
traditional I/Ki approach will not work that well
Therefore, it's
10
because if A affect B exposure, you have to
11
incorporate that dynamic interrelationship when you
12
do your predictive assessment.
13
Here, we use a PBPK approach prospectively
14
to not only assess a DDI risk but also use to
15
inform our phase 1B clinical trial design to see do
16
we need a single-agent lead-in phase for our
17
phase 1B combo trial.
18
you can see, the first molecule is a 3A4 and UGT
19
substrate, and at the same time is a 3A4 and 2D6
20
inhibitor based on the in vitro study.
21
new molecular entity is a 3A4 and 2D6 substrate and
22
is also a 3A inhibitor.
For the two molecules, as
The second
So here you can clearly
A Matter of Record (301) 890-4188
102
1
see, at least based on in vitro data, there are
2
bidrectional PK DDI risks in this case.
3
We did PBPK simulation and also assessing
4
multiple different dose levels at different and
5
type of combination, and it suggests that the
6
drug-drug interaction risk is low to moderate and
7
is actually low at the starting dose.
8
readouts support the strategy of starting at the
9
low dose for the first molecule in the combination
And these
10
studies and also without the requirement for
11
starting a single-agent lead-in phase for these
12
phase 1B combinations.
13
I can mention to you, for this specific
14
project, the project team was very nervous in the
15
beginning because to worry about the bidirectional
16
PK DDI between the two molecules, and especially
17
based on the earlier study from exposure safety
18
perspective, both molecules are very narrow
19
therapeutic window compounds.
20
So we want to ensure a very success start of
21
the phase 1B combination, and there's heavy vested
22
thinking whether we need a single-agent lead-in.
A Matter of Record (301) 890-4188
103
1
However, we know with the slow enrollment rate from
2
oncology patients, having a separate lead-in cohort
3
was prolonged, the duration, and also incurs the
4
cost of the trial.
5
So based on this simulation, we suggested
6
that it's okay, that we'll further refine the
7
assessment after we get data from the first cohort,
8
but it's okay to start at low risk.
9
retrospective observed data confirmed those
10 11
And the
findings. These are the PK examples, and then if we
12
get into the efficacy space, starting from the case
13
where we don't have any clinical data, what we can
14
do is base on xenograft data when we don't have any
15
clinical data.
16
Here's a diagram illustrating the
17
translation strategy we implemented at Genentech,
18
firstly, to build ad PK/PD relationship in
19
xenograft models, and then incorporate the human PK
20
totality -- use a human PK to correct for the
21
inter-species difference for drug exposure.
22
correct for protein binding and target binding as
A Matter of Record (301) 890-4188
You
104
1
needed and assume the same exposure relationship.
2
Based on these, you can predict what's the target
3
tumor growth inhibition at the clinical exposure.
4
How do we know that this translational
5
approach works?
Harvey Wong, our DMPK colleague,
6
actually did a very nice retrospective analysis
7
based on 8 anticancer agents.
8
was the detailed compounds listed on the left
9
table, and the right-hand side is showing the
And this analysis
10
correlation diagram.
11
very good correlation between the observed clinical
12
objective response on the Y scale and also the
13
predicted xenograft tumor growth inhibition driven
14
by human PK exposure on the X scale.
15
This analysis showed there's
This analysis suggests that more than
16
60 percent tumor growth inhibition in xenograft
17
models at clinical relevant exposure are most
18
likely to lead to clinical efficacy.
19
have been using this as a minimal efficacy target,
20
and we used 80 or 90 percent TGI targets, a more
21
optimal target.
22
In-house, we
I'm going to show you an example of applying
A Matter of Record (301) 890-4188
105
1
this translational approach to inform combination
2
therapy -- a new molecular entity combination study
3
the dose selection.
4
test the 2NME combos together.
5
schedules to check?
6
agents, we only have efficacy data in clinical from
7
phase 1 all-comer studies, so we really have
8
minimal clinical efficacy data available.
9
The challenge here is how to What are the dosing
Especially for both of the
The opportunity here is we do a step-wise
10
dose escalation trial design, targeting to have
11
with the molecules -- both molecules have different
12
degrees of tumor inhibition.
13
left-hand side is a typical step-wise dose
14
escalation when testing two molecular entities at
15
the same time.
16
molecules at cohort 1.
17
tolerability window, you simultaneously escalation
18
to cohort 2 and 3.
19
dose for one molecule.
20
then go into cohort 4, combining the two higher
21
doses of the two molecular entities.
22
Shown on the
You start with a low dose for both If it passes the
Each case, you escalate the If both are cleared, you
To inform the dose selection for the trial,
A Matter of Record (301) 890-4188
106
1
we used a translational PK/PD approach, as you can
2
see here, linking the PK of the MEK inhibitor in
3
blue, PI3K inhibitor in black, and linking these to
4
the tumor response in a additive fashion.
5
the xenograft data we have, it's suggestive that
6
additive is sufficient to describe the combo we
7
observed in xenograft models. Based on these analyses and the target, we
8 9
Based on
have learned based on the translational
10
strategy -- it tells us, as you can see, I color
11
coded based on if the dose combination will be
12
below the target, will be hitting the 60 percent
13
minimal target, or hitting the 80 percent optimal
14
target.
15
in the phase 1B studies have good probability of
16
anti-tumor activity -- have the good potential of
17
anti-tumor activity.
18
It suggests that the doses we're testing
That's a case where we don't have any
19
clinical efficacy data.
What to do when we already
20
have some clinical efficacy data, because the
21
clinical data always trumps the preclinical, some
22
of us say.
This is an example to show -- to
A Matter of Record (301) 890-4188
107
1
optimize dose and schedule for a combination with
2
standard of care.
3
tolerability profile we showed, it's caused a
4
question of whether we can potentially better
5
manage the tolerability profile with alternative
6
dose and schedule.
7
In this case, because of the
However, the challenging question we have is
8
do we have any potential risks of loss of efficacy
9
if we do intermittent dosing?
That gives us the
10
opportunity to see whether we can use longitudinal
11
tumor response modeling, for example, to optimize
12
in dose and regimen and also project what may
13
happen for the intermittent dosing without having
14
any observed data.
15
As you can see, the model diagram is very
16
similar to what we are doing for xenograft models.
17
Here this case, integrate the PK of the standard of
18
care in blue and the PK of PI3K inhibitor in black,
19
also in additive fashion, affecting the tumor
20
response, the only thing different will be for the
21
tumor response model here.
22
resistant factor incorporated, as typically we do
There will be a
A Matter of Record (301) 890-4188
108
1
for human tumor response model.
2
based on the data we have for PI stroke inhibitor,
3
it continues QD dosing.
4
we can make simulations to show scenarios for
5
intermittent dosing.
6
Using this model
After building the model,
Shown here are the predictions for what if
7
the standard of care alone in the yellow arm.
8
That's more like our virtual control arm, and also
9
the two greens are where we have observed data for
10
continuous QD dosing.
11
pink show the potential different intermittent
12
dosings.
13
And the blue, purple, and
Here, it shows, for the purple and pink
14
color, whether you are dosing 21/7 or 5/2 schedule,
15
you have a very low likelihood of loss of efficacy.
16
So before doing the intermittent trial study, this
17
gives us a comfort that you have a low likelihood
18
of loss of efficacy for intermittent dosing.
19
For efficacy, especially for combination,
20
it's a question of can you demonstrate target
21
inhibition in the combo setting, how to demonstrate
22
that pathway-specific inhibition.
A Matter of Record (301) 890-4188
The opportunity
109
1
here is whether we can do further dose
2
justification based on target-specific biomarker
3
responses and also PK/PD analysis.
4
This is an example from our AKT inhibitor
5
programs as shown on the pathway on the left-hand
6
side.
7
biomarkers, both at the systemic level shown in the
8
upper from the platelet enriched plasma and also
9
showing on the lower panel from the core tumor
For AKT, we collected two type of
10
biopsy.
11
the tumor biopsy biomarker, we show very good
12
target engagement.
13
From both the systemic biomarker and also
In addition, based on systemic biomarker,
14
you can see on the upper-right corner, therein lies
15
exposure-response relationship we have to write for
16
the AKT inhibitor, and it's supporting the dose
17
levels we tested in our multiple phase 2 trials,
18
including 200, 400, and 600 milligrams for our
19
phase 2's.
20
That's the efficacy aspect.
21
only half of the game.
22
importantly, is the tolerability.
Efficacy is
The other half, more
A Matter of Record (301) 890-4188
And I'm just
110
1
going to show you one example in the tolerability
2
space and more focused on how to benchmark, how to
3
compare with single-agent tolerability. The challenge we have in this case is how to
4 5
interpret the combo tolerability without a control
6
arm in the phase 1B study.
7
gives us is how to maximize the learning from
8
historical single-agent data, how we can integrate
9
this historical information to inform our ongoing
10 11
The opportunity that
trials. We did it based on two approaches.
One is
12
based on literature data base meta-analysis,
13
model-based meta-analysis.
14
are combining a PI3K inhibitor with paclitaxel.
15
do have in-house paclitaxel database that tells us
16
what's a typical efficacy and safety profile for
17
paclitaxel.
18
This is a case where we We
Then, as shown on the lower left-hand side,
19
the stars and triangles, these show observed
20
tolerability in our combination trial.
21
these observed tolerability events, incidence rates
22
are consistent in the ballpark with the observed
A Matter of Record (301) 890-4188
We can see
111
1
historical single-agent tolerability profile of
2
paclitaxel.
3
incidence rate of grade 3 and above.
4
at other AEs and other grades.
Here we focus on the neutropenia We also look
In addition, for neutropenia, there is a
5 6
very well established model for myelosuppression,
7
and here is a published -- the model has been well
8
published in the literature by Lena Friberg's lab
9
in Uppsala, especially for paclitaxel. Here, we are also using historically
10 11
published PK/PD model for paclitaxel on neutrophil
12
counts.
13
of the model prediction, based on historical PK/PD
14
model compared with observed neutrophil counts time
15
profile that we have from the combo study.
16
Different colors represent different dose levels of
17
PI3K.
18
And in the diagram showing the comparison
The totality of the readout showing that the
19
neutropenia incidence rate and also neutrophil
20
counts versus time profile in our combo trial are
21
consistent with the historical paclitaxel
22
monotherapy based on both the literature
A Matter of Record (301) 890-4188
112
1 2
meta-analysis and also PK/PD modeling exercise. These are just showing some of the
3
traditional PK/PD approach.
This is a slide I
4
borrowed, courtesy from my colleague, Saroja, who's
5
showing the potential opportunities for using
6
further quantitative system pharmacology approach.
7
Personally, I feel especially for oncology
8
combinations, this may be a unique area to give us
9
some opportunity for using the QSP because here, we
10
have a very multifaceted cross-over pathway, and
11
also you want to -- the QSP approach, it was a very
12
nice opportunity to assess what to combine in a
13
combination and also how to combine them
14
potentially using a more mechanism-based approach.
15
The quantitative system in pharmacology is
16
really more of also a data integration tool,
17
integrating data about the pathway, about the cell,
18
the tissue, the physiology, and pathology of the
19
patient, and also the patient heterogeneity.
20
area is a fast, evolving area, but it's still more
21
in the exploration stage.
22
work both in the academic field and also the
This
I think there is active
A Matter of Record (301) 890-4188
113
1
industry area to see whether this approach can
2
provide any value added for drug development.
3
this is an opportunity.
4
But
In summary, I hope my presentation today
5
gave you some flavor of what are some of the unique
6
challenges, and the correspondingly, what are some
7
of the unique opportunities this will provide us in
8
drug development?
9
[indiscernible] for trial design, how we can
The challenges are for
10
effectively and efficiently study drug combinations
11
for multiple dosing schedules.
12
Go back to the hot MDT topic from yesterday,
13
do we challenge the MTD paradigm because the
14
paradigm I show is still more of the traditional 3
15
by 3, testing for combination.
16
from preclinical to clinical, from early to late,
17
challenging for combination, and also for combo,
18
from PK, efficacy, and tolerability perspective.
19
The translation
This also gives us the opportunity from
20
study design perspective, for both preclinical and
21
clinical study, how we can smartly design a study,
22
give us opportunity for effective measurements,
A Matter of Record (301) 890-4188
114
1
what type of data to collect, the different
2
endpoints, how we can integrate them across
3
different molecules, different targets, to inform
4
the combo development.
5
Also, for modeling, as I'm a
6
pharmacometrician by training, data is golden.
But
7
hopefully, based on the data we have, we can also
8
have the opportunity for application of modeling
9
and simulation tools throughout the entire
10
development cycle to address for PK DDI, PI DDI,
11
and also benchmarking with historical data; is it
12
at the literature-published level or in-house
13
individual level?
14
I just want to highlight, that's not the end
15
of the story of 2NME combination because it's
16
getting more and more complicated, and many
17
companies are already doing that.
18
longer just two combinations.
19
highly poly combinations, triplicate, quadruplet.
20
I can hardly pronounce the word, frankly.
21 22
So it's no
There are very
Also, today's presentation is focused on combination of two small molecule entities to align
A Matter of Record (301) 890-4188
115
1
with a workshop topic.
2
combination of multiple molecule types, small and
3
large, small and ADC, large and ADC.
4
combinations provide, also again, very unique
5
challenges and opportunities in drug development of
6
how to do those very smartly.
7
However, we are doing
All these
Also, the immuno-oncologist is such a hot
8
evolving area, so now today's talk more focuses on
9
combination of kinase inhibitors.
But then we are
10
also doing combination of multiple novel
11
mechanisms.
12
with a kinase inhibitor, combination of multiple
13
immunotherapies.
14
considerations there?
15
What about the combined immunotherapy
What are some of the
Also, for combinations, it goes beyond the
16
dose level and regimen.
17
sequence?
18
dose titration, as shown by a very nice example by
19
Yazdi, that if, for example, for combination, if
20
you have a dose titration, how to handle that in a
21
combo setting.
22
What about dosing
Which one do you dose first?
And also,
The story really goes on, and I'm hoping my
A Matter of Record (301) 890-4188
116
1
presentation today is just going to hopefully
2
trigger more discussions.
3
saying that we throw a stone, hoping to get a jade
4
back.
5
during the panel discussion from this evolving
6
topic.
7
There's a very famous
So I'm hoping that we'll get a lot of jade
I cannot end without thanking all my
8
colleagues and collaborators, especially at
9
Genentech and also all the clinical investigators,
10
the animals, and also the patients.
11
colleagues, Mark Dresser and Bert Lum, are in the
12
audience with me today.
13
your attention.
14 15 16
Two of my
So thank you very much for
(Applause.) Panel Discussion - Qi Liu and Julie Bullock DR. BULLOCK:
Okay.
We're going to move to
17
the panel discussion portion of the meeting for
18
Session 1, so if the panelists could please come up
19
and take your seats.
20
While they're getting seated, I'll go ahead
21
and introduce myself.
I'm Julie Bullock.
22
director of clinical pharmacology for D3 medicine,
A Matter of Record (301) 890-4188
I am
117
1
and I'm also a 10-year ex-FDA survivor.
2
left in November, so if I still talk like a
3
regulator, it's because I haven't quite been
4
brainwashed completely yet.
5
I recently
I would like the people that have not been
6
either on a panel or haven't spoken yet to
7
introduce themselves so that people know your
8
affiliation and your background.
9
DR. BELLO:
Tunde?
My name is Tunde Bello.
I'm the
10
executive director for clinical pharmacology and
11
immuno-oncology at Bristol-Myers Squibb.
12
DR. BULLOCK:
13
DR. RATAIN:
Mark? I'm Mark Ratain.
I'm a medical
14
oncologist and clinical pharmacologist at the
15
University of Chicago.
16
DR. BULLOCK:
Vivek, or Atiqur?
17
DR. RAHMAN: I'm Atiqur Rahman.
I'm the
18
division director of Division V in the U.S. Food
19
and Drug Administration, which supports the
20
oncology and hematology drug product development.
21 22
DR. KADAMBI:
I'm Vivek Kadambi.
I'm the VP
of nonclinical development at Blueprint Medicines.
A Matter of Record (301) 890-4188
118
DR. BULLOCK:
1
All right.
We have an hour, I
2
promise this will be hopefully a productive and
3
fruitful moderation and discussion.
4
burning question, though, as I was reading the bio.
5
My first question's for Dan Howard because I
6
noticed that you spent a lot of time in Kansas
7
City, which is where I grew up.
8
where your favorite barbeque place is.
10
DR. BULLOCK:
11
DR. HOWARD:
14
DR. BULLOCK: more questions, then.
16
(Laughter.)
17
DR. BULLOCK:
19
I know.
barbeque. (Laughter.)
18
Hard question.
Actually, I don't like
13
15
So I must know
(Laughter.)
9
12
I first have a
Well, I'm not asking you any I'm done.
Your opinion doesn't matter to
me. Anyways, no, I really do have a real serious
20
question for you.
I really appreciated your
21
lessons learned that you stated.
22
slide, but I wrote them all down because I think
A Matter of Record (301) 890-4188
You didn't have a
119
1
that they are really spot-on.
But when you're
2
going through your example, I feel like what you
3
are able to do at Novartis would not have been
4
possible without the fact that you had robust PK in
5
your program all the way through.
6
really telling because I think there's a lot of
7
opportunity in pharmacokinetics, but yes, there's
8
also a lot of challenges.
And I think it's
What were some of the logistical issues with
9 10
this?
11
there any resistance that you thought was different
12
between getting PK in an oncology testing versus
13
your experience that you had previously in drug
14
development that was a non-oncology?
15
always hear that we can't do this in oncology.
16
Did you test multiple cycles for PK, and was
DR. HOWARD:
Wow!
Because we
What a loaded question.
17
Well, first of all, there wasn't any resistance for
18
the LDK or the ceritinib program primarily because,
19
frankly, the team knew very well that we were on to
20
something very exciting, and they didn't want to
21
do -- they didn't want to minimize the importance
22
of the things that we needed in order to go fast.
A Matter of Record (301) 890-4188
120
1
When it came time to make a selection, for
2
instance, about what we were going to do in terms
3
of pharmacokinetics, we had single dose, we had
4
steady-state pharmacokinetics, and we also had a
5
substantial collection of sparse data to support
6
population analyses.
7
Keep in mind that a lot of the decisions
8
that were made for this particular product were
9
made in a very, very short period of time, so when
10
a debate came up about what we should do or what we
11
needed, it was often supported as this is the
12
justification we need in order to provide for the
13
registration of this compound fast.
14
stopped all debate.
That usually
15
Now, as far as the second part of your
16
question, is it different between oncology and
17
non-oncology, and do I find it more difficult or
18
less difficult, I'll be honest with you.
19
came over to the department three years ago, I was
20
struck by the fact that in oncology, we were
21
choosing to collect data in a way that it didn't
22
provide for answering the kinds of questions I
A Matter of Record (301) 890-4188
When I
121
1 2
wanted to answer with development. For instance, we might collect -- as any of
3
you are probably aware, in our Farydak submission,
4
we collected pharmacokinetic information in the
5
study that had no efficacy, and we collected no
6
pharmacokinetic information in the study that had
7
efficacy.
8
between this information.
9
where we were collecting subset analyses, so we
I was not able to make a connection We have other programs
10
would collect 10 subjects, for instance, out of a
11
group of 250.
12
Now, that doesn't really provide much in the
13
way of connecting exposure and response either for
14
safety or for efficacy.
15
I've been there, we've changed this so that we are
16
now collecting it in all of ours studies in a
17
manner that allows for an exposure-response
18
analysis across the board.
19
DR. BULLOCK:
In the three years since
Yes.
I think that's really
20
important.
I think a lot of the FDA people are
21
nodding their heads, the clinical pharmacologists,
22
at the limitations of subset analyses and not
A Matter of Record (301) 890-4188
122
1
getting PK in efficacy trials.
I've known drugs
2
that have been held up because they didn't have
3
that information.
4
unfortunate that many -- especially smaller
5
companies I feel haven't really pushed for it and
6
looking more at pharmacokinetics.
7
pain is the fifth vital sign.
8
that PK is the sixth.
9
considering more PK than some of these arbitrary
It's very valuable, and it's
They always say
Well, I would say
So we should probably be
10
vital sign stuff that we get throughout development
11
all the time.
12 13 14
Does anyone else have any comments on this type of things?
Mark?
DR. RATAIN:
Go ahead.
Yes.
I hear Dan's concern
15
about my colleagues in oncology, their perspective
16
on pharmacokinetics.
17
to do a biopsy to find out why the patient didn't
18
respond.
19
checking a plasma level of the drug?"
20
answer is, "Well, why would we want to do that?"
21
So there is too little use of pharmacokinetics, and
22
to be frank, overuse of biopsies to understand
You often hear, well, we need
And I go, "Well, have you considered
A Matter of Record (301) 890-4188
And the
123
1
clinical pharmacology of modern oncology drugs.
2
DR. BULLOCK:
Yes.
I couldn't agree more.
3
I have just another question.
I was sitting
4
on the train for a long amount of time this
5
morning, so I got to think a little bit about
6
yesterday's sessions.
7
know if it's because I work with a lot of the
8
smaller clients now in my new role, and we had the
9
great presentation by Dr. Bailey yesterday on
What came to me -- I don't
10
different types of Bayesian methods for first-
11
in-human trials. But I think what I'd like to get you guys'
12 13
opinion on is maybe the 3 plus 3 design isn't
14
broken.
15
fact that we're still using cytotoxic DLT criteria
16
to define DLT.
17
getting enough good data and looking at it,
18
including PK and PD.
19
expanding it more than one dose.
20
because we're using cancer patients in situations
21
when a biomarker might be present in a healthy
22
volunteer, and therefore it might be better to go
Maybe what's broken about 3 plus 3 is the
Maybe it's because we're not
Maybe it's because we're not
A Matter of Record (301) 890-4188
And maybe it's
124
1
to a healthy volunteer first to look at these PK
2
biomarker things.
3 4 5
So does anyone have any thoughts on my morning train, uncaffeinated revelations? DR. MASSON:
I think the 3 plus 3 is not
6
necessarily broken, but I think it's the
7
flexibility we build in the protocol that allows
8
the dose escalating quickly.
9
start very slowly for some compound because it's
Sometimes we have to
10
not been previously tested.
11
single-patient dose escalation or titrate until we
12
start seeing toxicity, and then we can expand.
13
So we may have to do a
So I think it's especially important in
14
phase 1 that we characterize the safety target
15
engagement in PK.
16
efficacy, but that's rare, unless you have very
17
targeted agents and you do your expansion solely in
18
that tumor type.
19
characterize also the dose response for efficacy,
20
how much is this applicable to other tumor types?
21 22
Sometimes you get lucky; you get
But that begs a question.
That's one of the challenges.
If you
We might pick
a dose in one tumor type that might not be the
A Matter of Record (301) 890-4188
125
1
optimal dose in the other tumor type that may have
2
different mutation status and may not need to be
3
hit this hard.
4
think, is the expansion, because typically when we
5
have a drug, we'll look at signals in several tumor
6
types.
7
we'll do dose response where we see signal.
8
that's one of the challenges for the dose response.
11
So we'll do several expansions, and then
DR. BULLOCK:
9 10
So that's one of the challenges, I
Okay.
So
Anybody else on the
panel? DR. BELLO:
I think, clearly, we're dealing
12
with pressures from both sides, really, in terms of
13
maybe a level of conservatism from our
14
investigators and their familiarity and comfort
15
with the 3 plus 3.
16
concerns from our drug development organizations
17
around the operational -- not to mention just the
18
operational challenges, but the interpretation of
19
the data from the newer model based approaches.
20
Plus also, sometimes internal
So I think Julie's point around if we have
21
more data, if we have more examples, if we show the
22
utility of these approaches, then that's likely to
A Matter of Record (301) 890-4188
126
1 2
move the needle with regard to broader examples. DR. BAILEY:
For myself, I think the
3
challenge of the 3 plus 3, it's not just 3 plus 3
4
but an algorithmic design that relies on one piece
5
of data to decide whether to go up or down.
6
just relying on one piece of data.
7
of information you actually collect, there could be
8
many different reasons why you would choose to
9
remain where you are, even if you've not seen
10 11
You're
With the wealth
toxicity. DR. RUBIN:
And the small number of the
12
patients.
13
typically, it's a pretty heterogeneous group, and
14
you could enroll a couple of outliers in a 3 plus
15
3, and you're way off track.
16
I mean, these phase 1 oncology trials,
DR. BAILEY:
Yes.
I think the concept of
17
the sample size being small, 3 patients per cohort,
18
I have no issue with, really, cohort sizes being 3,
19
4, 6, 1, it's fine.
20
follow a design that makes a decision based on this
21
one piece.
22
You can make it more conservative, but it's still
But it's the concept that you
And you can define DLT how you like.
A Matter of Record (301) 890-4188
127
1
relying purely on the toxicity, and that's now how
2
you do effective dose finding. DR. RATAIN:
3
I think that the most important
4
issue that would really be nice to come to terms
5
with is what is the purpose of a phase 1 trial.
6
the purpose to define a single dose for subsequent
7
studies?
8
that.
9
of the data.
10 11
If so, okay, fine.
We know how to do
And at the end of the day, you look at all However you got there doesn't matter
so much; that's just a matter of efficiency. But to be frank, I don't think that's what
12
we should be doing going forward.
13
should be defining and understanding the
14
relationship between exposure and toxicity.
15
really should be the primary purpose.
16
phase 1 trials should be conducted under, in the
17
case of oral drugs, perennial conditions that
18
maximize exposure, maximize bioavailability.
19
Is
I think we
That
I think
It doesn't make sense to me to do phase 1
20
studies under conditions of low bioavailability if
21
the purpose is to understand relationship of
22
exposure to toxicity.
And I think we have to take
A Matter of Record (301) 890-4188
128
1
any exposure efficacy relationships with a large
2
grain of salt.
3
pharmacodynamics will vary from disease to disease.
As pointed out, the
DR. KADAMBI:
4
If I may add a question for
5
Stuart and Dan.
On ceritinib, the breakthrough
6
therapy status, the great data that was generated,
7
would that have been different on a 3 plus 3
8
design?
9
it proved itself well.
You did it using your Bayesian method, and That's the question out
10
here, is it broken not to be fixed?
11
the analysis.
12
worse?
13
Go back and do
Would that be indifferent, better,
DR. BAILEY:
So in terms of the escalation,
14
I need to go back and check, but I don't believe we
15
would have ended up with the high dose had we been
16
using a 3 plus 3 purely because of an intermediate
17
observation of DLT that would have forced us back.
18
The collection of PK data would be no different.
19
The challenge would have been to be able to get the
20
expanded cohort sizes, the expanded patient
21
numbers, to be able to have more confidence in that
22
PK data at a variety of dose levels.
A Matter of Record (301) 890-4188
129
1
DR. HOWARD:
Just to throw something out, as
2
Alice pointed out to me -- Alice Shaw pointed out
3
to me -- we probably couldn't have gone to 900 no
4
matter how I wanted it.
5
felt comfortable that we had explored a pretty wide
6
range of effective doses, and it was really not
7
useful to go any higher.
8
design have changed that?
9
that by the time we reached that 750, everybody had
10
agreed, we won't really want to go any higher if we
11
don't have to.
12
The investigators really
So would a 3 plus 3 I doubt it.
I think
Now, what I said to Alice at the first break
13
was maybe what we should have done was possibly go
14
a higher exposure, by then exposing somebody to
15
food plus ceritinib and measured it that way.
16
might have been an alternative.
17
hindsight's 20/20, that's probably what I would
18
have done.
19
DR. BULLOCK:
20
DR. JIN:
That
And I think if
Jin?
I agree with Dr. Ratain that the
21
main objective for a phase 1 study is really
22
not -- it's really the objective of the study.
A Matter of Record (301) 890-4188
If
130
1
we identify the objective, not to identify the MTD,
2
but really to have an effective phase 2 dose, or
3
even doses to test in the phase 2, that's really
4
the objective.
5
For Julie, I agree with your comment because
6
I bear the same question yesterday during the
7
discussion.
8
for the design.
9
we do at the end of phase 1, to really integrate
To me, it's actually not as critical It's really what type of analysis
10
the data we collected, the information we have.
11
The PK, efficacy, safety biomarker, response
12
everything, whether it's from a 3 by 3 design or
13
adaptive design, it is whether we collect the data
14
more effectively and also more efficiently.
15
I agree that the early cohort, if we start a
16
very low dose, there will be a timing benefit or go
17
through those initial dose cohorts quickly.
18
However, from a dose response or exposure-response
19
perspective, it's almost like the only time we're
20
collecting any information from the low-dose
21
patient during the phase 1 escalation.
22
study, you only have an opportunity to refine at
A Matter of Record (301) 890-4188
At later
131
1
your target dose or target exposure.
2
make a difference if you skip some of the early
3
small doses with just one or two subjects rather
4
than wait and do 3.
5
dose you're getting I think will be case by case.
How much gain and how much
Audience Q&A
6
DR. BULLOCK:
7 8
So will that
from the audience. DR. MORCOS:
9
I'd like to take a question Go ahead.
Pete?
Peter Morcos with clinical
10
pharmacology at Roche.
On the ceritinib example,
11
could another lesson learned be to just leverage
12
the physicochemical properties right on and just go
13
directly to fed dosing right from the beginning
14
with a poorly soluble poorly permeable drug? Secondly, the dose optimization worked for
15 16
ceritinib as a PMR.
17
back and then try to optimize afterwards.
18
these studies then designed as new pivotal trials?
19
Are they BE studies?
20
actually do with the data that you generate in this
21
study?
22
It's an interesting way to go But are
And what are you going to
The third question is on these accelerated
A Matter of Record (301) 890-4188
132
1
study designs, where you enroll N equals 1, N
2
equals 2, to try to get to MTD.
3
exclude PK from your analysis because how do you
4
then evaluate the dose exposure relationship?
You almost then
For example, the example that no PK increase
5 6
with increasing dose, how confident are you in that
7
or is that just variability due to close dose
8
increment increases, et cetera?
9
PK in your assessment of phase 2 dose selection
How do you include
10
with very small sample sizes using these
11
accelerated escalation schemes?
12
So three questions.
13
DR. BULLOCK:
14
DR. HOWARD:
15
(Laughter.)
16
DR. HOWARD:
Three questions. Yes, possibly, and no.
Let me say yes.
The first
17
part, there are lots of lessons to pick out of
18
this, not the least of which is the one you
19
summarized.
20
though.
21
study where we're looking at whether or not
22
tolerance is improved and we maintain our
Let me go to this second question,
What do we do with the data from the PMR
A Matter of Record (301) 890-4188
133
1
benefit-risk ratio by targeting the exposure given
2
with a meal?
3
Well, what we're going to do with
4
that -- you can call it pivotal, you can call it
5
whatever you like.
6
study which we will then go back to the agency, and
7
we will discuss whether or not this information
8
should be incorporated into the label, and how it
9
should be applied.
But actually, it will be a
It is very likely that if
10
indeed the food does allow us to increase GI
11
tolerability, then we definitely want to get that
12
in the label because it's good for patients.
13
I'm sure the agency will agree.
And
14
DR. MORCOS:
As dosing change?
15
DR. HOWARD:
I'm sorry?
16
DR. MORCOS:
As a dosing change to that
DR. HOWARD:
I can imagine that's exactly
17 18
extent?
19
what will happen.
20
talks about how we change dose, depending on the
21
safety profile.
22
DR. BELLO:
We'll see maybe a section that
Again, from a perspective of
A Matter of Record (301) 890-4188
134
1
someone who relatively recently worked on the
2
crizotinib program that was -DR. HOWARD:
3 4 5
It's a good program, by the
way. DR. BELLO:
-- thank you -- that ceritinib
6
followed up on, I think some of the issues that you
7
actually met and tried to address were things that
8
we encountered and addressed in our program.
9
again, crizotinib didn't get breakthrough therapy,
And
10
but was approved under accelerated approval.
And
11
the development program was very rapid, and we had
12
to move extremely quickly, as similarly to you.
13
We noticed the GI tolerability issues, and
14
one of the approaches -- the half-life of the drug
15
actually does support once-daily dosing, but the
16
drug is dosed BID predominantly based on the
17
observation of GI tolerability.
18
potentially another means of actually increasing
19
your overall exposure, but not necessarily
20
increasing that GI tolerability effect.
21 22
So this was
Again, I wonder a little bit about some of the -- and maybe there will be discussions around
A Matter of Record (301) 890-4188
135
1
this if we skip around a little bit -- decisions we
2
make are based on maybe product profile from a
3
commercial perspective that maybe override a little
4
bit the science and what we could be doing to kind
5
of actually address some of these issues in another
6
way.
7
DR. HOWARD:
That's a very good point.
8
me just say one thing.
9
possibility of a BID dosing.
Let
We did think about the But at the time that
10
we entered our trial, we had a 2-hour before and 2-
11
hour window after where we were restricting food
12
intake because we were really uncertain with
13
this -- at first -- BCS4, how much increase we
14
might see.
15
after, BID dosing kind of became a little
16
difficult.
17 18 19 20 21 22
With a 2-hour before and a 2-hour
DR. MORCOS:
More of a reason to maybe just
go directly into fed. DR. RATAIN:
So did you consider just doing
your phase 1 with food rather than fasting? DR. MORCOS: asking initially.
Yes.
That's the question I was
So if you knew potentially of
A Matter of Record (301) 890-4188
136
1
BCS4 increased exposure with food, why not just go
2
directly as fed straight from the beginning?
3 4
DR. HOWARD: Lilli Petruzzelli?
5
(Laughter.)
6
DR. HOWARD:
7 8 9 10
Could somebody hand a mic to
She's shaking her head.
She
says, yes, we did consider giving it with food. By the way, this is Lilli Petruzzelli, everyone. DR. PETRUZZELLI:
I oversaw the study once
11
it entered clinic, so some of those decisions were
12
made and we were already dosing in clinic.
13
actually entertained whether to switch and add food
14
in the middle.
15
food study at the time we were beginning to see
16
efficacy.
17
in switching.
18
But we went ahead with the formal
It was something we strongly considered
The initial decision I can't comment on why
19
they were so strict.
20
food early in our studies.
21
DR. HOWARD:
22
And we
We are doing more and more
Well, let me say this.
Actually, for me, the real question of whether you
A Matter of Record (301) 890-4188
137
1
start with food or without food, or frankly, even
2
if you dose with food or without, it is a question
3
of, in some respect, bioavailability.
4
it's more a question of variability.
5
that we can control the variability by giving the
6
drug with food, we're not increasing the
7
variability, then we know that we're bringing some
8
consistency to the patients that should result in
9
consistent safety and efficacy profiles.
But I think If we know
Frankly speaking, I would rather have low
10 11
bioavailability with low variability, a regimen,
12
than I would have a regimen that increases the
13
bioavailability but ends up with a high degree of
14
variability.
15
DR. PETRUZZELLI:
16
DR. RATAIN:
Especially -- yes.
Let me respond to that, Dan,
17
because we've actually studied that and published a
18
paper.
19
variability goes way up, interindividual
20
variability.
21
agency, with the exception of oncology, in regard
22
to labeling drugs with positive food effects to be
Under low bioavailability conditions,
And there's consistency in the
A Matter of Record (301) 890-4188
138
1
taken with food.
2
for reasons that escape me, an editorial by Larry
3
Lesko, may he rest in peace, said something to the
4
effect that cancer patients have trouble eating,
5
which of course isn't true, particularly for
6
chronic diseases like chronic myelogenous leukemia.
7
And it's the oncology -- again,
DR. HOWARD:
It sounds tempting to say we
8
should have started with food in this.
But you
9
know, honestly, I wouldn't change what we did.
We
10
would have started in fasting anyway.
11
preclinical data indicated that our total
12
bioavailability, at least in our animals, was
13
somewhere between 30 and 60 percent.
14
a whole lot of uncertainty regarding whether it was
15
a moderate bioavailability or a low
16
bioavailability.
17
that we could control the drug before we introduced
18
another variable.
19
Our
So there was
We wanted to get in and make sure
DR. BULLOCK:
This is a great conversation,
20
but I'd like to move on just a little bit because
21
we only have about 30 minutes left.
22
Pete, did you get your questions answered
A Matter of Record (301) 890-4188
139
1
for the most part? DR. MORCOS:
2 3
Yes, the first two.
The third,
I know there's a bit of time -- all right.
4
DR. BULLOCK:
5
(Laughter.)
6
DR. BULLOCK:
7
DR. YATES:
Okay.
Thanks.
Some other time.
Next? Hi.
I'm James Yates,
8
preclinical modeling and simulation, AstraZeneca.
9
I did have two questions, but I've been here a
10
while, so I've got a third now.
11
(Laughter.)
12
DR. YATES:
Sorry.
But the first one's quite short.
13
So the ceritinib example -- sorry; another question
14
about that -- you were comparing preclinical and
15
clinical exposure.
16
protein binding differences?
17
DR. HOWARD:
18
DR. YATES:
19 20
Did you correct for plasma It wasn't clear.
Yes. Okay, good.
Told you it was
short. The second one, it's around axitinib, but
21
also some of the other examples where it was
22
obvious that you were going to start at one dose,
A Matter of Record (301) 890-4188
140
1
and then go up or down.
2
analysis highlight any covariates that would have
3
enabled you to start at another dose?
4
DR. PITHAVALA:
Did the population PK
Yes, this question comes up
5
all the time, is there any demographic or
6
pharmacogenomic, any other characteristics which
7
might identify the patients who are going to dose
8
increase or decrease.
9
looked at everything under the sun, demographic
We looked at everything.
We
10
factors.
11
where we looked at pharmacogenomics for the CYPs,
12
for transporters, et cetera, and none of those
13
items found any covariance.
14
if you are of this ethnicity and this body weight,
15
you're more likely to dose titrate, but we looked
16
and we didn't find anything.
17
We did a huge combined 500-patient study,
DR. YATES:
Okay.
We would love to have,
Thanks.
The third one's
18
for the whole panel.
Maybe Jin Jin can answer
19
first, though.
20
preclinical models to generate that translational
21
relationship.
22
cell line.
It's around the choice of
Most examples tend to use only one
I wonder what the panel thinks about
A Matter of Record (301) 890-4188
141
1
that, whether we should be using a range of models.
2
The idea of mouse clinical trials, using a range of
3
PDX models to generate some variability in the
4
response, et cetera.
5
DR. JIN:
I think at least from our internal
6
experience, there's almost never one case.
We only
7
have one xenograft model.
8
number of xenograft models we have probably range
9
from 20 to maybe even 100, especially for
It's generally -- the
10
combination with different submutation types or
11
resistance mechanism or certain things prior
12
treated.
13
When we do the translation, of course one
14
way is you integrate the data from multiple
15
xenograft models to have some kind of range of
16
uncertainty.
17
look at by scenarios, what about some optimistic
18
models, what about some kinds of rigid models?
19
Also, most importantly, what is the most
20
mechanistic base relevant to the clinical
21
situation?
22
Another case, a quicker way is we
DR. YATES:
Any other thoughts?
A Matter of Record (301) 890-4188
Thanks for
142
1 2
that. DR. BELLO:
Yes.
Maybe another one is
3
around -- again, clearly you need to utilize
4
multiple cell lines and also I think a BMS.
5
There's an approach to actually incorporate
6
patient-derived cell lines not from long held or
7
many generations of the same cell lines to make
8
sure that you're trying to get almost, as it were,
9
a real patient, real world kind of assessment of
10
your drug.
11
DR. BULLOCK:
12
DR. DE ALWIS:
Go ahead. Dinesh De Alwis, quantitative
13
pharmacology and pharmacometrics, Merck.
14
actually have two questions.
15
interesting talks all around.
16
the other --
17
DR. HOWARD:
18
(Laughter.)
19
DR. DE ALWIS:
I
By the way, very One is to Dan, and
Is this about ceritinib?
Yes, it's about -- this is
20
what happens when you have an excellent
21
presentation.
22
had clinical data, and really nice clinical data,
I actually wanted to understand, you
A Matter of Record (301) 890-4188
143
1
showing that lower efficacy worked, Kaplan-Meier
2
curves.
3
preclinical data and this way showed that higher
4
doses or high exposures are better because of the
5
resistance and you're trying to overcome that.
6
But yet, you kind of went back to the
I'm trying to understand.
I'm obviously an
7
advocate of translational approaches, but there
8
comes a time when you're in the clinic, and
9
clinical data probably trumps the preclinical data.
10
Just your thoughts on why you continued to explore
11
the higher doses.
12
DR. HOWARD:
I'm glad you brought that up.
13
I mentioned something early on when I was going
14
over some of the nonclinical information that we
15
were using to help us guide the dose selection.
16
There was one thing about ceritinib that we hoped
17
to exploit to benefit patients, and that was its
18
ability to cross the blood-brain barrier.
19
it had some ability to cross the blood-brain
20
barrier, it meant that there was a possibility for
21
patients that we might affect brain mets.
22
it only benefited us at the very highest exposures
A Matter of Record (301) 890-4188
Because
Again,
144
1
because it doesn't have a tremendous ability to
2
cross, but it has some.
3
factor in the decision-making.
4
DR. DE ALWIS:
5
DR. HOWARD:
6
DR. DE ALWIS:
And that was another major
Thanks. Sure. The next question is to
7
Yazdi, and it was, towards the end of your
8
presentation, you showed a ORR I think of around
9
60 percent, but the PFS -- I know it was empowered
10
for PFS, but it was a hazard ratio 0.85.
11
wondering the disconnect there.
12
always correlate well with survival and PFS.
13
you look at change in tumor size?
14
understand that a little bit better.
15
DR. PITHAVALA:
I'm just
RR [ph] doesn't Did
And I'll
Yes, it's hard.
It's hard
16
to get a definitive answer.
17
even with the small number of subjects, to see a
18
clear difference in the PFS, but we did not.
19
Again, as I mentioned during my presentation, it
20
was a very small subset of patients.
21 22
We would have liked,
In general, at least in RCC, we're in the process of developing a model which shows very
A Matter of Record (301) 890-4188
145
1
clearly that early tumor shrinkage seems to predict
2
the PFS down the road.
3
disconnect between those two endpoints, again, in
4
second-line RCC.
5
types; certainly lung cancer, and a similar model
6
has been developed in CRC as well.
7
So I don't think it's a
I can't speak to other tumor
So I don't think it's a disconnect issue.
8
really do believe it's a sample size issue.
9
small group of patients.
10
DR. DE ALWIS:
Yes.
It's a
Because obviously the
11
issue is the power between RR versus looking at
12
something continuous.
13
correlation better.
14
information? DR. PITHAVALA:
16
DR. DE ALWIS:
17
in tumor size and PFS.
18
DR. PITHAVALA:
19
DR. DE ALWIS:
20
DR. PITHAVALA:
22
So I'm just curious was the
You don't have that
15
21
No. Between tumor size -- change
In that particular study? Yes. I don't believe we've looked
at that. DR. DE ALWIS:
I
Okay.
Thanks.
A Matter of Record (301) 890-4188
146
1 2
DR. PITHAVALA:
But it's worthwhile to look
at, for sure.
3
DR. ROY:
Amit Roy from BMS.
4
three questions, but --
5
DR. BULLOCK:
6
DR. ROY:
7
I don't have
Good.
-- but I do have one comment and
two questions.
8
(Laughter.)
9
DR. ROY:
Actually, the first comment, I
10
wanted to go back to Julie, what you opened with,
11
the idea of collecting PK in efficacy studies.
12
think it's a very valuable thing to do.
13
said, it can -- I'd like to caution if it's not
14
then well, it can actually lead you in the wrong
15
direction.
16
exposure is only as good as the data that you
17
collect in your trial.
18
I
That being
Your characterization of PK and
I just want to make the case for collecting,
19
first of all, informative PK samples.
20
are a number of tools that are actually out there
21
that allow you to collect informative PK samples.
22
Also, bear in mind while doing this that there can
A Matter of Record (301) 890-4188
And there
147
1
be two objectives of collecting PK samples.
2
to characterize the PK In a patient population, and
3
the other is to characterize the estimate, the
4
exposure of a given individual.
5
sampling scheme that you would actually implement
6
is not necessarily the same for both of those two
7
things.
8 9
One is
And the kind of PK
So for example -- and it's quite well recognized -- if you just take random PK samples
10
across the dosing interval, that will give you
11
generally a good characterization of PK in the
12
patient population.
13
samples from an individual, that's not going to
14
give you a good estimate of your overall exposure
15
because oftentimes your trough -- so you need to
16
kind of bear those two things in mind.
17
But if you only have peak
The third thing that I wanted to mention
18
is -- well, actually, there are two more with
19
regard to PK sampling.
20
we found, especially with the dasatinib example
21
that I talked about yesterday, it's useful, when
22
you go down this informative sampling thing,
For small molecule drugs,
A Matter of Record (301) 890-4188
148
1
recognizing that you have interoccasion
2
variability.
3
So if you're just taking PK samples on one
4
occasion, you're likely to get another true
5
representation of the exposure of a drug,
6
especially if there's a lot of interoccasion
7
variability to bioavailability, for example.
8
it's advisable to take it multiple occasions.
9
other one is, in order to do this well, you need to
10
have a very good PK data cleaning kind of operation
11
because otherwise, a small number of samples in a
12
patient, then you could get widely incorrect
13
estimates of exposure.
So The
14
So that's the one comment on a number --
15
DR. BULLOCK:
No.
It's a very valid
16
comment, and I think it's very important that it
17
came up in this setting because I agree that the
18
data is only as good as the data that you collect.
19
And if you collect crappy data, you can't torture
20
anything out of it.
21
getting nothing, but there is something worse.
22
It's getting something that doesn't make any sense.
There's nothing worse than
A Matter of Record (301) 890-4188
149
1
DR. ROY:
Right.
2
DR. BULLOCK:
And that's one of the problems
3
with subset analyses and bad sampling schemes.
I
4
think my favorite one when I was at the agency was
5
someone was like, "We're getting PK.
6
trough."
7
hours.
8
like you're sampling for nothing."
9
really excited about that.
10
(Laughter.)
11
DR. BULLOCK:
We're getting
And I was like, "Your half-life's 2 You're not going to have anything pre-dose; But they were
But the other thing I think
12
that is important to understand within oncology is
13
because we have so many dose reductions, and dose
14
delays, and dose interruptions, that cycle 1 PK is
15
not informative for the rest of the treatment
16
cycle.
17
into cycle 1 only, and then they forget about the
18
rest of treatment.
19
missed opportunity.
20 21 22
And a lot of companies will incorporate PK
And I think that that's a huge
Any other comments on this topic or we can have Amit have his other question? DR. ROY:
Yes.
The second question should
A Matter of Record (301) 890-4188
150
1
be short.
2
vandetanib example.
I wanted to give Howard a
3
break a little bit.
So you had a dose -- you were
4
testing a lower dose of post-approval of the
5
300 milligram.
6
respect to enrollment of patients at that lower
7
dose?
Did you have any issues with
DR. PITHAVALA:
8 9
This is actually for Eric and the
That's a good -- it is
difficult to enroll that population.
So the only
10
thing I can say is it was an 81-patient trial.
11
took two years to enroll, complete the study.
12
it was done at 20 sites across the globe.
13
don't think it impedes -- because clearly we
14
had -- we were expecting efficacy at 150 as well.
15
So I don't think that was an issue, but I know
16
Kevin is in the room, actually, and is responsible
17
for the program.
20
And
So I
Kevin, do you have any other thought on
18 19
It
this? DR. HELLER:
Since we were already approved
21
in the United States, a lot of the sites were in
22
Europe because we did not get approved in Europe.
A Matter of Record (301) 890-4188
151
1
So there was that point.
2
study, which of course if we did that here after
3
approval, it still would have been very hard
4
because no one would want a 50 percent chance of
5
getting a lower dose.
6
promise of crossover after progression or when we
7
do the unblinding.
8 9
And it was a double-blind
But also, there was the
I think we designed it in order to maximize, ethically, an opportunity for patients to cross
10
over if there's benefit.
11
outside of the United States, there was no other
12
way they could have gotten access to the drug.
13
DR. ROY:
And also, by doing it
The last question I have,
14
actually, is for the entire panel.
15
speakers have alluded to using continuous measures
16
of tumor burden to assess response.
17
may be useful to have a bit more discussion about
18
that because a lot of decision-making still occurs
19
based upon RECIST response rates.
20
I made yesterday was that in RECIST response -- or
21
the RECIST criteria were intended to really I think
22
make sure that the change in tumor size was
A Matter of Record (301) 890-4188
A number of
I thought it
And a point that
152
1
actually robust in the sense that it was not
2
correlated to efficacy, but certainly if you had
3
more than 30 percent decrease in tumor size, that
4
was a real decrease in tumor size, and more than
5
20 percent is probably an increase in tumor size.
6
Something in between, not quite sure.
7
So what's used, and very appropriately, for
8
making a decision about treatment for an individual
9
patient may not be the most appropriate thing to
10
make a decision in drug development to move a
11
compound forward.
12
with using longitudinal data -- because you have
13
data from multiple patients that you're making this
14
decision on, on drug development program, that
15
might be a better way to go in the future rather
16
than using just simply response rates.
17
And maybe a continuous measure
DR. MASSON:
I think we should explore that,
18
especially for immuno-oncology, where you don't
19
necessarily see great overall response, but yet you
20
see tumor regression that's prolonged.
21
we need to look at that carefully.
22
think we need to look at new technology like
A Matter of Record (301) 890-4188
So I think
And also, I
153
1
circulating tumor DNA, and see if that's a better
2
correlate to long-term efficacy.
3
DR. HOWARD:
There's no question about the
4
fact that you want to use continuous variables when
5
you can, but in the end, everybody's going to drive
6
you back to telling you what the PK/PD or what the
7
exposure-response results are for my OS, my PFS, my
8
response rate, my DOR, all of that.
9
DR. PITHAVALA:
The only other comment about
10
using the continuous ones is it will work in almost
11
every instance, but you need to validate it.
12
need the tumor type because if you have a tumor
13
type where a lot of the progressions are because of
14
new lesions popping up and not the target
15
lesions -- there are many instances when your
16
target lesions might be shrinking, but your
17
progression, you meet your PFS endpoint because a
18
new lesion popped up.
19
the model, unless you have prognostic factors which
20
are predicting for which patients will have new
21
lesions pop up.
22
DR. BAILEY:
You
That won't get picked up in
This is a recent challenge we
A Matter of Record (301) 890-4188
154
1
had.
We actually had a phase 1 trial that was
2
designed with a primary endpoint of exposure tumor
3
response as the primary, with the DLT being
4
monitored as a key secondary, and exactly the point
5
of new lesions coming up and not being able to
6
continue the estimation.
7
DR. ROY:
If I could just make a follow-up
8
point on that, I think what's been recognized in
9
immuno-oncology is oftentimes patient benefit in
10
being continued on the treatment even though new
11
lesions do pop up.
12
overall tumor burden does decrease.
13
some modified criteria that have been applied
14
successfully.
15
wouldn't work for target therapies.
16
overall tumor burden is shrinking and you have a
17
new lesion pop up, it's not clear that that might
18
not go away with continued treatment.
19
Ultimately, this sort of So there's
It's not clear that the same thing
DR. RUBIN:
If your
Yes, that's right.
And my
20
understanding is that that was done with
21
crizotinib, is that there was an allowance for
22
people to stay on treatment even though they met
A Matter of Record (301) 890-4188
155
1
the progression criteria because of that concern.
2
I think you're right.
3
50 percent -- if your overall tumor burden goes
4
from 30 centimeters down to 10 centimeters, but
5
you've got a tiny new lesion, why would you stop
6
treating that patient?
7
I'll just say in the area of immunotherapy, it's
8
become even more of a concern since these things
9
could represent inflammatory reaction to a small
10 11
If you've got a
It makes no sense.
And
tumor site. DR. JIN:
And also, longitudinal tumor PK/PD
12
I think provides an opportunity to refine exposure
13
response, estimate the addition to the objective
14
response PFS as we typically look at.
15
much additional benefit is adding the PK/PD
16
relative to the traditional exposure response?
17
least internally at Genentech, I think this is an
18
area we are still actively investigating, is that
19
truly more beneficial?
20
However, how
At
So that's where actually I'm interested to
21
learn from other fellow companies and also
22
regulators experience.
One thing we do feel,
A Matter of Record (301) 890-4188
156
1
there's a unique benefit for the longitudinal
2
model, and it's actually to address a regimen
3
question because it's actually predicting
4
alternative regimen or different dosing schedules.
5
Looking at the analysis in a longitudinal fashion
6
provides opportunity to predict the response in a
7
longitudinal fashion, where the traditional
8
exposure response for OR, or PFS, or OS, those
9
things cannot be addressed.
10
DR. BULLOCK:
11
DR. HABTEMARIAM:
So just my point.
Bahru, go ahead. My name is Bahru
12
Habtemariam.
13
My first question is to Eric.
14
particularly mixed results for the dose
15
optimization trial, where the lower dose showed
16
better tolerated, better reduced efficacy.
17
there some sort of modeling exercise or something
18
to use some intermediate -- was some intermediate
19
dose considered other than the 150?
20
I'm one of the reviewers at the FDA.
DR. MASSON:
So you showed us
Was
I think dose in between are too
21
small to get benefit.
So again -- I think
22
85 percent tolerated without dose reduction of 150
A Matter of Record (301) 890-4188
157
1
versus 71 percent.
2
receive 300, but if they have toxicity, then they
3
need to reduce.
4
is not possible for this compound given the long
5
half-life.
6
observed late and also some of the toxicity, you
7
can really use early indicator of safety or any
8
other biomarkers to predict efficacy.
9
So that means the majority can
Unfortunately, I think titration
And the fact that the efficacy is
Yes, I think still the 300 I think is the
10
optimal dose, but we did look at exposure response
11
from that study as well, and it was consistent, at
12
least for the QT based on the previous trial.
13
DR. HABTEMARIAM:
My second question is
14
basically for everyone.
It looks like now there is
15
some consensus that estimating MTD doesn't make
16
sense for non-cytotoxic toxic drugs because the old
17
cytotoxic drugs are toxic to begin with, so you
18
have to get [indiscernible] to give it to patients.
19
So moving forward, should the primary objective of
20
the phase 1 oncology trials be activated estimation
21
and how the build the mechanism for stopping?
22
if that's not the case, is there any -- what's
A Matter of Record (301) 890-4188
Or
158
1
limiting us in terms of making the transition? So I would like everybody to comment if you
2 3
can. DR. BAILEY:
4
Maybe I can comment to that.
I
5
don't think it's right to say we shouldn't think
6
about estimating MTD.
7
in general, safe -- I mean, we've clearly seen
8
examples where even with the small molecules, you
9
end up with an MTD or the highest kind of tolerable
10
Just because the doses are,
dose. I think the key is to be able to have the
11 12
flexibility, as you say, to not make that the
13
absolute goal.
14
biomarker driven endpoint.
15
recently done, a tumor response exposure endpoint
16
to look for minimum effective dose, but then use
17
the safety to define a range.
The primary endpoint could be a It could be, as we have
So you're actually trying to look that up,
18 19
and it was actually for an antibody, so it's more a
20
cost of goods issue to be able to keep the dose
21
low.
22
understanding the properties of your drug and
But I think that, as we suggest,
A Matter of Record (301) 890-4188
159
1
understanding the disease that we're going into and
2
how those two are connected is really going to tell
3
you the key question to prioritize for that
4
treatment.
5
question, but it's a phase 1 trial, and patient
6
safety has to be maintained.
7
And the safety may be a secondary
DR. HABTEMARIAM:
Sure, but if you're
8
monitoring safety, you will determine MTD by
9
default.
Right?
10
DR. KADAMBI:
Not if you don't -- not if
11
you're increasing the dose.
12
DR. BAILEY:
13
DR. KADAMBI:
14
DR. BAILEY:
If you increase the dose. Right. We had a discussion yesterday
15
about this, the concept of how accurate do you have
16
to be on your MTD estimation.
17
to give the MTD, you want to be accurate on that
18
estimation.
19
without any toxicity and you start to show quite a
20
wide therapeutic index, you maybe don't need to be
21
so accurate on the MTD estimation because you've
22
already shown this window of safety for covering
If you end up having
If you are able to see high activity
A Matter of Record (301) 890-4188
160
1
any DDI potential and other dosing errors that may
2
incur.
3
where you're never going to look.
4
something there, if you need to be there, to be
5
able to accurately assess it.
6
So it's not that you're in a situation You have to have
To Vivek's comment earlier on about the
7
design, yes, there is an issue with 3 plus 3.
I'm
8
not necessarily backing the 3 plus 3 because if
9
you've only got one or two doses you're looking at,
10
it's not so bad.
11
you would have got there, but in general, the
12
longer the dose range you study, the less likely
13
you are to get to high doses with the 3 plus 3.
14
Just to Eric's point, 3 random, and that's the big
15
challenge.
16
But the case for ceritinib, maybe
DR. KADAMBI:
Actually, if I may make a
17
comment on that?
I think on the targeted therapy,
18
for example, if you have a covalent inhibitor that
19
specifically modifies your target -- and looking to
20
MTD actually doesn't make sense.
21
can demonstrate your receptor is occupied -- I
22
mean, the challenge to say how your
A Matter of Record (301) 890-4188
As long as you
161
1
receptor -- there are a lot of covalent inhibitors
2
that are being developed.
3
does not have to get those, so I think it's a valid
4
point.
5
protein that you modify is your protein of
6
interest.
7
exaggeration of pharmacology there.
8 9
I think the MTD concept
Hopefully, the target therapy, the only
So hopefully, the toxicity is an
DR. RATAIN:
Stuart, you were using the word
"accurate," and I think you meant precise.
And
10
there's a huge difference.
11
believe when they're writing a precise dose,
12
182 milligrams per meter squared, and the DSA is
13
1.7635, then the pharmacists and the nurses recheck
14
the math that were precisely giving an accurate
15
dose to the patient.
16
I think oncologists
I think we need to get away from the whole
17
concept of precision given interindividual
18
pharmacokinetic variability and focus on the big
19
picture.
20
an MTD I think for a population, but it's more
21
important to understand the characteristics to
22
determine inter-individual variability, and we lose
And I agree, we want to accurately define
A Matter of Record (301) 890-4188
162
1 2
sight of that. DR. BAILEY:
Yes.
My meaning is more around
3
the level of confidence that you have that this is
4
an acceptable dose, and the level of confidence you
5
have that the dose you studied above is not.
6
DR. JIN:
In my opinion, actually, I feel
7
MTD has its unique value, understanding the
8
tolerability, even for target agents, at least in
9
the cadre of kinase inhibitors because these
10
molecules generally still have relatively narrow
11
therapeutic window.
12
they still have a lot of tolerability issue.
13
They're wider than chemo, but
The MTD by nature, maximum target dose,
14
actually tells a lot of information of what's the
15
maximum actually tolerated exposure because this
16
actually tells you the information when you really
17
dose a drug in the patient population, especially
18
either with potential PK drug interactions or in
19
special populations where some patients may have
20
high exposure.
21
exposure potentially?
22
that, you need the information.
What is that highest tolerated Since actually MTD provides
A Matter of Record (301) 890-4188
163
I think the limitation where we are is not
1 2
MTD per se.
3
shouldn't be bound of considering MTD as our
4
phase 2 dose.
5
dose to move forward in the later clinical
6
development.
7
the right dose, but it's just some other cases MTD
8
is not the right dose and is basically not bound by
9
that.
10
It's basically from the phase 1.
We
It's basically what is the right
In some cases it may be that MTD is
That's my opinion. DR. BELLO:
I absolutely agree with that.
11
think at the end of the day, MTD -- the right dose
12
is going to be closer to the MTD if it's a narrow
13
therapeutic index drug.
14
a wide therapeutic index and you've got a high
15
confidence in understanding mechanism of action,
16
and your biomarkers tie up and run together, then
17
clearly it makes sense not to dose to MTD.
18
I
If you've got a drug with
But I think, again, we're in a way preaching
19
to the choir in terms of us folks being the folks
20
that really understand and are thinking about how
21
to better come to a dose.
22
some mechanical, operational, or historical, or
But I think there are
A Matter of Record (301) 890-4188
164
1
traditional issues within our organizations in
2
terms of drug development.
3
But also in terms of the external
4
investigator community around, okay, what's the
5
fastest path for me getting to a dose that has some
6
level of a high level of probability of being
7
effective and then being able to move the program
8
relatively quickly?
9
have, again, this level of history and confidence
I think the investigators
10
in the MTD approach.
11
challenge to come back to them and say, okay, we
12
have to have the confidence and the though process
13
in terms of deriving the package and the arguments
14
that support other approaches.
15
think the ultimate challenge here.
16
DR. KADAMBI:
And it's really, again, a
And that's really I
If I may make one point, I
17
think maybe it's a native notion here.
We talk
18
about targeted agents on one hand.
19
some ways it's an oxymoron if the MTD does not turn
20
out to be exaggerated pharmacology then it's not a
21
targeted agent in a sense.
22
envision to say I have a targeted agent, and yet my
The MTD, in
I mean, how can one
A Matter of Record (301) 890-4188
165
1
MTD is not exaggerated pharmacology?
2
not targeted.
3
And then it's
So conceptually, I think we need to
4
understand with a notion of targeted.
5
naive point, but --
6
DR. HOWARD:
Maybe it's a
Even a targeted agent -- I
7
mean, you can have effects that are maybe not
8
directly related to your target inhibition.
9
effects for instance, we've had a number of
GI
10
different things that limit how far we can go with
11
the dose.
12
DR. KADAMBI:
But that could be, as you
13
mentioned, Stuart -- I mean, not Stuart, Dan, that
14
this could be a local effect of your drug.
15
agree with you.
16
attribute that is the chemistry of the molecule,
17
not a pharmacology of the target modulation.
18
And I
That's a physical chemical
DR. HOWARD:
I just want to say one thing.
19
Actually, it's not about MTD so much as it's about
20
therapeutic index.
21
therapeutic index in your phase 1 trial, you may
22
not need to go to MTD.
If you can define the
In fact, you won't want to
A Matter of Record (301) 890-4188
166
1
go to MTD.
And this is why large molecules don't
2
do it.
3
twofold above what I know is a therapeutically
4
targetable dose, therefore I don't need to go any
5
further.
They're able to say, look, I can get to
DR. BAILEY:
6
There are specific on-target
7
toxicities that will define potentially MTD anyway.
8
I worked on a compound, the on-target toxicity of
9
24-hour shutdown of the pathway led to ocular
10
opacity and soft tissue mineralization.
So you're
11
trying there to get into the range of a dose
12
whereby you have very, very low risks of those
13
kinds of toxicities. The other toxicities like the GI tox, you're
14 15
willing to go to another standard.
16
then, about trying to identify maybe a change of
17
schedule or intermittent dosing to give you a
18
window of dosing where you start to see on-target
19
effects that are not as severe as that, but then
20
leading you into this range. To Mark's point, it's not necessarily the
21 22
But it varies,
MTD.
If we talk about what's the goal of the
A Matter of Record (301) 890-4188
167
1
trial, it is to identify a therapeutic index window
2
for this compound, and then from that to be able to
3
select the dose.
4
DR. KADAMBI:
Fair point.
I think in that
5
case -- my last comment before Karthik has a
6
question he's waiting for.
7
point, Dan, then the therapeutic index, my default
8
is 1, right?
9
toxicity is equal to pharmacology.
In that case, and your
There is no therapeutic index if the Then all you're
10
trying to modulate is the schedule to be able to
11
get tumor efficacy in the absence of normal tissue
12
toxicity.
13
DR. BULLOCK:
14
that you would like to ask.
15
DR. LIU:
Yes.
Qi, I think you had a question
I just want to make one
16
comment in response to Dr. Mark Ratain's comment
17
earlier on food effect in oncology.
18
say the agency's actually very open-minded to
19
feedback from the community.
20
reason I intentionally showed two case studies,
21
olaparib and ceritinib, where we did not give the
22
drug under fasting condition.
I just want to
And that is the
A Matter of Record (301) 890-4188
We recommended
168
1 2
something different. Actually, as a matter of fact, science is
3
involving -- so is regulatory science.
Our office
4
is considering revising the food effect guidance.
5
And our deputy division director, Dr. Brian Booth,
6
is the co-chair for this working group.
7
sure they will try to address food effect oncology
8
patients.
9
workshop on food effect, and if you have any
And I'm
Several months ago, they had a public
10
additional questions, comments, or suggestions,
11
please feel free to let us know.
12
DR. BELLO:
Thank you.
Maybe I'll just add a little bit
13
to that.
14
food effect workshop and, again, talking to this
15
issue around utilization.
16
straightforward as everyone thinks in terms of,
17
okay, you administer a meal and you talk about high
18
fat, low fat.
19
I was an attendee and panelist at that
It's maybe not as
There are so many variables that also have
20
to be considered with regard to meals.
Just to
21
name a few:
22
differences, specifications and expectations of
ethnic differences, food type
A Matter of Record (301) 890-4188
169
1
uniform performance of your drug under those
2
conditions.
3
depending on your drug's behavior with food,
4
exactly what this individual's supposed to eat
5
maybe for the rest of their life.
6
You're almost specifying, in
a way,
This is a consideration as we talk about
7
incorporating food effect into our program.
8
it's clearly something that needs to be
9
investigated, but it's not as straightforward as
10 11
So
one might think. DR. RATAIN:
But there are many examples
12
outside of oncology.
13
telaprevir's label, for example, TID, the meal
14
should contain approximately 20 grams of fat, and
15
the label gives instructions for what that means.
16
Look at a Novartis compound, Novartis drug,
17
Coartem, which has I think the largest food effect
18
that exists in the pharmacopeia, where it makes it
19
very clear that for treatment of individuals of 5
20
kilograms or more with severe malaria -- a pretty
21
sick group -- that you have to give it with food.
22
I mean, if you look at
So I mean, everybody does this outside
A Matter of Record (301) 890-4188
170
1
oncology, and there are parts of the world where
2
there's very little technology even.
3
we should be able to figure out how to do it for
4
our patients with our modern drugs.
5
DR. PITHAVALA:
And I think
I think regardless of
6
whether you do a [indiscernible] study, you start
7
with a fasted state or fed state, I think it's
8
probably a good idea that once you get into the
9
dose expansion phase with whatever dose you're
10
moving forward, to do a pilot assessment of food
11
right there in that study.
12
forward, you have that information.
13
DR. BAILEY:
So as you're moving
Can I just necessarily
14
challenge to do that in the expansion phase because
15
from the experiences that we've had, often if you
16
do end up with an MTD and then try doing it, and
17
you have some information about whether you think
18
it may or may not be positive or negative, it
19
doesn't necessarily translate.
20
actually then introduce undue risk if you're doing
21
it in the expansion.
22
And you can
It's better to do it earlier to
A Matter of Record (301) 890-4188
171
1
understand --
2
DR. PITHAVALA:
3
DR. BAILEY:
Absolutely.
-- maybe one or more doses, and
4
then hopefully you can get to the expansion, even
5
without the limitation of food, and that makes the
6
development even easier.
7
DR. BULLOCK:
Let's move on a little bit.
8
Food effect's important.
We've got it.
And
9
there's going to be a workshop, even better.
10
Before, when we were discussing this, and I
11
was discussing stuff with Qi, we had some questions
12
that we wanted to ask the panel.
13
I sent out on a survey to have blinded results
14
because they were a little bit tree-shaky in the
15
terms of like who should be involved in the dosing
16
decision within companies.
17
these types of things?
18
kind of sensitive.
19
And some of them
Who should be driving
And I knew that they were
The one thing is that when all of these came
20
back, everyone said the exact same thing for who
21
should be involved in the dosing decision.
22
clinical, pharmacology, pharmacometrics, the
A Matter of Record (301) 890-4188
It was
172
1
clinician, the biostatistician, as well as
2
pharmacology, toxicology.
3
responses said the same five things.
4
commercial.
5
no one had formulations in that as a discipline.
6
Every single one of the No one had
No one had regulatory, no one had CMC,
So I was wondering if anyone would like to
7
make any comments on why these disciplines that you
8
chose were important or why you think the other
9
ones are less important.
And I'd also like to talk
10
a little bit about formulation availability and the
11
ability to have available formulations for dose
12
modifications, which I think is huge, and it limits
13
clinical trials.
14
Would anyone like to take this discussion?
15
DR. JIN:
I guess the reason for those
16
functions being selected, as a scientist by nature,
17
for dose selection, dose optimization, it's
18
fundamentally first as a data and science-driven
19
decision.
20
like clinical operation, like the regulator has,
21
these many times are logistic aspects, and the
22
operation of those are very important.
And for other aspects like formulation,
A Matter of Record (301) 890-4188
But the
173
1
initial selection, the decision, recommendation
2
should more come from a data and science base. Then after having that at least initial
3 4
recommendation, we can figure out the logistic
5
operations and maybe hopefully try to be creative
6
to have a path forward.
7
my opinion as a scientist by nature.
8 9 10
At least, I think that's
DR. BULLOCK:
That's great, Jin.
DR. KADAMBI:
I agree on that.
Thank you.
Stuart? I think one
11
of the things we need to not forget -- and I think
12
you said that, Julie -- is formulation, CMC.
13
takes about six months to get something going from
14
the time a clinician says I need a 5-milligram, or
15
a 10-milligram, or a 25-milligram dosage strength.
16
We should not underestimate that, traditionally,
17
when you do your first-in-human, you probably have
18
3 dosage strands, your starting dose, which is
19
fixed, and then you have some multiple, and your
20
maximum dose that fits in a capsule without a
21
triple size zero or something of that nature.
22
It
So I think in some ways, maybe we were naive
A Matter of Record (301) 890-4188
174
1
not to include those functions.
And I agree, it's
2
not operational.
3
They would feel offended if we called
4
operational.
5
dosage strength that's between 10 and 100, but we
6
won't know for a while, and what is the minimum
7
time you need.
8
months as long as you have API available to do
9
that.
I think it's still a science. them
Include them to say we may need a
And I think the answer's always six
I think it's a very important thing we
10
should not dismiss because that can halt a trial
11
midstream.
12
DR. BELLO:
I completely agree.
I think,
13
again, from one of the presentations earlier today,
14
we had recommendations around -- and I think
15
again -- ceritinib, trying to make sure that you
16
have your commercial formulation as soon as
17
possible and utilize that within your clinical
18
development program because a lot of curve balls
19
can be thrown by formulation changes, unexpected
20
changes.
21 22
Again, the number of manufacturing sites all add to complexity and all add to the potential
A Matter of Record (301) 890-4188
175
1
areas of error or variation, so clearly, having a
2
formulation that's close to commercial.
3
challenge clearly is, with a breakthrough therapy,
4
the accelerated development time line, pharm sci
5
colleagues, and the CMC colleagues being able to
6
generate these final formulations.
7
you need to have a very strong collaboration and
8
partnership with them.
9
DR. BAILEY:
But the
So definitely,
To the point about the team
10
making a decision, I think the first thing I would
11
say is that one group that maybe was missing -- and
12
maybe I misheard -- was the operations group that
13
actually run the study, working with vendors,
14
working with sites.
15
the meeting making a decision; it's about coming
16
together at the time patients are actually being
17
recruited to make a decision about what data you
18
actually want to use in the next meeting because
19
the logistics of actually getting that data in to
20
be able to use it are actually quite complex.
It's not just about being in
21
I showed this wonderful paradigm.
22
taken 10 years to get to this paradigm where you
A Matter of Record (301) 890-4188
It's
176
1
can actually have this kind of thing.
2
simple as just we've got data, therefore analyze
3
it.
4
background.
5
It's not as
There's a big component to that in the
DR. BULLOCK:
Yes.
There's definitely a
6
theme of time in a lot of the responses.
7
lack of time to analyze data, collect data, and
8
look at data to make a decision sometimes.
9
think clinical operations is huge and having that
10 11
There's
So I
to be a seamless process is very important. We are out of time, unfortunately.
12
time for lunch.
13
up here.
It is
I've got this blinking red light
It's making me panicky.
14
(Laughter.)
15
DR. BULLOCK:
So if you have any questions
16
that you would like to direct to the panel, you can
17
just catch them during lunch or during the breaks.
18
But thank you all for a great first session this
19
morning.
20
(Applause.)
21
(Whereupon, at 11:47 a.m., a lunch recess
22
was taken.)
A Matter of Record (301) 890-4188
177
A F T E R N O O N
1
S E S S I O N
2
(1:01 p.m.)
3
Presentation - Geoffrey Kim DR. KIM:
4
I think we're all gathering
5
together post-lunch.
6
lunch.
Hopefully, you had a nice
7
My name is Geoffrey Kim.
I'm the director
8
of the Division of Oncology Products 1.
9
those that aren't aware, Oncology Products 1 does
So for
10
breast cancer, gynecological cancers, gender
11
urinary cancers, and supportive care.
12
briefly just introduce the session, and then
13
Dr. Sridhara's going to follow up with the
14
feasibility of the integrative adaptive
15
dose-finding trials.
16
to talk about barriers to implementing integrative
17
adapted dose-finding trials.
18
break.
19
I'm going to
Then Dr. Petruzzelli is going
And then we'll have a
Like I said, good afternoon ,and thank you
20
for everyone who has stuck through this two-day
21
workshop so far.
22
really interested in this topic, so that's great.
If you're here, that means you're
A Matter of Record (301) 890-4188
178
1
Subjectively speaking, as I'm a workshop organizer,
2
I'd say this has been a great interdisciplinary
3
discussion so far.
4
focus right now to putting it all together and
5
address how to implement these best practices in
6
small molecule kinase inhibitor drug development.
7
This has been somewhat alluded to before.
But we really want to shift our
8
But some of these kinase inhibitors are associated
9
with a high rate of dose reductions and
10
discontinuations in clinical trials.
But the real
11
concern is that out in clinical practice, this may
12
translate to poor adherence to therapy in the real
13
world, where there is less motivation and
14
accountability to comply with the prescribed
15
treatment regimen.
16
So for even blockbuster or breakthrough
17
therapy, there are many more issues in the real
18
world that may reduce patient compliance, such as
19
financial considerations, comorbidities, and gaps
20
in patient education and communication, on top of
21
the toxicity associated with treatment.
22
So this highlights the importance of
A Matter of Record (301) 890-4188
179
1
identifying key inefficiencies of the entire
2
development process so that during the entire life
3
cycle of the product's development, key issues such
4
as dose optimization are continually being
5
addressed in the hopes of providing real-world
6
benefit out in the community.
7
Not being privy to the behind-the-scenes
8
processes in the pharmaceutical companies, I have a
9
few very naive observations regarding the
10
development process.
11
appears to be that there is room for more
12
interdisciplinary communication.
13
sometimes it really seems like y'all ain't even
14
talking to each other amongst the disciplines.
15
The first thing is there
That is to say
We heard yesterday some very stellar
16
examples regarding the need to have close,
17
interdisciplinary communication for successful
18
development because it also appears that data and
19
findings from each discipline can provide critical
20
information throughout the entire life cycle of
21
drug development.
22
move away from the concept that during this phase,
And it would be important to
A Matter of Record (301) 890-4188
180
1
we only listen to nonclinical, but then afterward,
2
we put them back in a corner and take the
3
pharmacometricians out of the box.
4
forward toward a more integrative approach, all
5
disciplines should really be present and accounted
6
for in order to really integrate the data.
7
As we move
With that being said, it's quite apparent
8
that industry has far more data regarding
9
interdisciplinary analyses than what is submitted
10
to regulatory authorities and that industry has
11
already adopted an integrative type of approach.
12
And in the past two days, we have heard several
13
best practices to that approach.
14
So I want to make a plug here that sometimes
15
it may be in your best interest in a regulatory
16
application to include more information regarding
17
your approach, as it does clarify things on our end
18
to see how emerging data has shaped the development
19
of the product at the time of the review because,
20
after all, it is always a review issue.
21
(Laughter.)
22
DR. KIM:
My final observation would be that
A Matter of Record (301) 890-4188
181
1
if collectively we as critical stakeholders in this
2
drug development do not start learning from our
3
prior experiences, we will not address the key
4
inefficiencies of the development, and we'll
5
continue to struggle with dose optimization. So in terms of learning about our past, here
6 7
it is.
This is pulled from a 2012 paper from
8
Rosenberg and DeVita, chronicling key developments
9
in oncology.
And here we are right here.
So after
10
the surgical era, and after the chemotherapy era,
11
here we stand.
12
So what era do we define ourselves?
Are we
13
in the kinase inhibitor era?
In my script, I said
14
this has really been tasked.
I think most people
15
would agree that we've kind of moved on.
16
we are in the ceritinib era, given the attention
17
that we've been putting to that drug development.
18
So if we're not in the kinase inhibitor era, are we
19
in the genomic era?
20
would say, "You're so late.
21
he's actually said that.
22
But maybe
So to quote Dr. Pazdur, as he You're so 2008."
What about other omics, proteomics,
A Matter of Record (301) 890-4188
And
182
1
metabolomics, other types of omics and putting
2
together?
3
should we just all pack it up and go home because
4
we're clearly in the immunotherapy era, and none of
5
this is relevant anymore?
6
the best way to describe this era is that we are in
7
the informatics era.
8 9
Is that really the era that we're in or
But I think, clearly,
Yesterday, I had a really major snafu with my FDA ID badge, and I couldn't access my network
10
because of that.
11
around trying to my connections restored so I could
12
send these slides in, it became clear to me that
13
without access to this information that we share
14
with each other, and without a proper data
15
infrastructure, all this rapid development would
16
not be possible.
17
So as I was frantically running
So can you imagine going back to relying on
18
archived journals in the library sending budding
19
wanna-be med students to go in and copy the
20
journals with a copy card, and xeroxing all those
21
things, and hand it back to the PI to get the
22
information that we really need?
A Matter of Record (301) 890-4188
I don't think it
183
1
would be possible to speed the accelerated pace of
2
development that we have without having access to
3
reliable information.
4
So how do we capitalize on being part of the
5
information era?
One of the problems we face is
6
that there is so much data coming in from multiple
7
data sources at an extremely rapid pace.
8
comes in all different formats, and it's very
9
difficult to ascribe the proper weight to the data
This data
10
and filter the signals from the noise.
11
often extremely difficult to integrate these data
12
together because of the different formats and
13
different weights to the values of the data to
14
perform meaningful analyses.
15
day, these data really are all integrated and
16
interrelated, and potentially can be quite useful
17
if we step back and look from a global perspective.
18
And it's
But at the end of the
So really, in the ideal drug development
19
program, what we like to see is full integration of
20
meaningful data throughout the entire life span of
21
a product.
22
profiling of the molecule in nonclinical and
This includes full pharmacological
A Matter of Record (301) 890-4188
184
1
toxicology studies; intensive pharmacometric
2
profiling and modeling based on drug exposure; and
3
novel clinical trial designs that will integrate
4
historical clinical experience and key nonclinical
5
and pharmacologic data to allow for accurate
6
estimations of the adverse reaction profile of the
7
drug.
8 9
At the end of the day, sometimes you have to start out with what you want to figure out how to
10
get there.
11
discussion is the implicit understanding that what
12
we really are trying to do is get highly effective
13
drugs to the patients, which is certainly in line
14
with what they are looking for.
15
patients are also telling us -- and we know this
16
through a series of patient-focused drug
17
development workshops that we have been having at
18
the agency -- is that they want to have a realistic
19
expectation of what taking any particular drug will
20
do to them.
21 22
I believe that inherent to all our
But what the
They don't want to be surprised.
Are we at a point where we as developers can start to put together this collected data so that
A Matter of Record (301) 890-4188
185
1
instead of simply describing what has happened in
2
clinical trials involving patients who may or may
3
not be in the same age, demographic, ethnic,
4
genomic, or other category as them, we can switch
5
to a paradigm of providing an educated estimation
6
of what their experience will be like based on who
7
they are.
8 9
So yes, this all sounds like pie in the sky, too much Kool-Aid kind of stuff here.
But if it
10
doesn't happen, it certainly will not be from the
11
lack of trying.
12
start things off, but this is the type of systems
13
pharmacology analysis we are starting to piece
14
together.
15
more and potentially collaborate with a lab to
16
explore the concept of endothelial and cell
17
dysfunction as it relates to kinase inhibition and
18
vascular occlusion.
19
So Dr. Simpson presented this to
And certainly, we would like to explore
Dr. Kluwe yesterday presented a powerful
20
example of an animal model of hypertension and
21
prevention of left ventricular dysfunction with an
22
ACE inhibitor but not a beta blocker.
A Matter of Record (301) 890-4188
Along those
186
1
same lines, understanding the mechanisms by which
2
vascular occlusion occurs with these kinase
3
inhibitors may be important, as there may be
4
concomitant medications that mitigate the risk,
5
such as a long-acting nitrate should this be
6
related to vasospasm.
7
But it will not stop there, and we are
8
fortunate enough to have internally a very talented
9
staff at the FDA with wide-ranging skills such as
10
Liang and Mang [ph], who are looking into the
11
development of a Bayesian confidence propagation
12
neural network that can reliably present the
13
associations of kinase inhibition to adverse
14
reactions.
15
To take it even one step further, Jiang
16
Liu [ph]
and James Zhou are working on integrating
17
population PK data into the mix to see if a kinase
18
association score can be established.
19
then is to take these associations and to see if
20
they can be factored into a Bayesian type of dose
21
escalation.
22
conceivably be very useful for the double and
The idea
If this holds, this type of data could
A Matter of Record (301) 890-4188
187
1 2
triple combinations that are being developed. Sorry.
This is exactly the kind of work
3
we're trying to do here, but it's still in the
4
process of being developed.
5
Someone said, while we were developing this
6
workshop, that FDA needs to say that they don't
7
like the 3 plus 3.
8
the 3 plus 3.
9 10
So here it is.
We don't like
(Laughter.) DR. KIM:
But we also do recognize that it
11
won't go away completely.
12
situations where it can and will be useful.
13
also want to move away from the 20-day DLT
14
definition, as it really appears to lead to an
15
overestimation of the overall tolerance of a drug
16
or combination.
17
even for breakthrough products.
18
adoption of best practices, including novel
19
approaches to data integration.
20
There are certain But we
We like to see more doses studied, And we encourage
We would also like to see data continuously
21
being integrated to the entire life span of the
22
product, including the postmarketing setting,
A Matter of Record (301) 890-4188
188
1
because at the end of the day, this is really about
2
producing a better product to the patient in the
3
most efficient manner possible.
4
So this room and the telephone line is full
5
of the unsung heroes in the development process.
6
I'm not sure if I ever heard a patient say, "Boy.
7
I'm really grateful for the discovery team for
8
de-risking this compound before pursuing further
9
development."
Or, "Boy.
I'm sure glad that a
10
pharmacometric team optimized this dose through a
11
thorough exposure-response analysis."
12
deserve so much credit for truly transforming the
13
way people with cancer are being treated in this
14
era we live in.
15
collective vision and continuously improving of how
16
to get this done.
17
But you all
And I think we all share a
So doing this session, we really want to get
18
at the barriers to implementing these strategies so
19
that we can be aware of how to improve on the
20
current system.
21
will really happen.
22
inefficiencies in the system and optimize this
Also, we want to discuss how this How can we reduce
A Matter of Record (301) 890-4188
189
1
process?
2
us all how to get this done.
Presentation - Rajeshwari Sridhara
4
DR. SRIDHARA:
5
I don't think I have all the
answers.
7
(Laughter.)
8
DR. SRIDHARA:
9
Thank you so much.
(Applause.)
3
6
So Raji is here, and she's going to tell
But let's see what we heard
in the last three sessions.
So I'll start off with
10
the current product development, how it is right
11
now.
12
have a preclinical, a phase 1, a phase 2, a
13
phase 3, and then back in phase 4 dose finding.
14
preclinical, we assess DLTs within a 28-day cycle,
15
cumulative toxicity unknown.
16
phase 1, but we determine the so-called MTD.
17
heard a lot of discussions on that.
18
the phase 2, use that MTD, multiple cycles of
19
treatment, and frequent dose modifications we start
20
seeing there; and use some soft efficacy endpoints,
21
ORR.
22
It says if it's working in silos, that you
So
The same thing with We
And then in
Then we go to phase 3, and we continue to
A Matter of Record (301) 890-4188
190
1
use the same MTD or maybe a little modified MTD;
2
multiple cycles of treatment, and again, frequent
3
dose modifications.
4
endpoint of OS.
5
because there were so many dose interruptions or
6
dose modifications.
But here we have a clinical
And then dose finding back again
So why is this so?
7
Because we are using an
8
outdated paradigm, and that's the cytotoxic
9
paradigm.
It worked very well for cytotoxics.
The
10
products were being given in 28-day cycles, finite
11
number of treatment cycles; dose based on BSA,
12
which was a substitute for exposure-based dosing.
13
Let's give that there.
14
time.
15
models; well characterized toxicities; and we have
16
the CTCAE grading criteria to get all of it, and
17
DLT defined based on these toxicities.
18
Toxicity observed in short
More is better is the model; good animal
However, the current products, the examples
19
that we are talking of kinase inhibitors, they're,
20
in general, oral formulations of fixed doses
21
administered beyond the 28-day 6 cycles until
22
disease progression every day; cumulative toxicity
A Matter of Record (301) 890-4188
191
1
and delayed toxicity that is observed in
2
preclinical or dose-finding studies that are not
3
observed; and type of toxicities are very different
4
as well.
5
Then we have to really ask the questions, do
6
we have adequate animal model?
Maybe not.
Current
7
study design unable to predict some toxicities.
8
Yesterday, we had very good discussion of some of
9
these, like the ILDs that cannot be predicted using
10
some of the animal models that we have.
11
have to really question is the definition of the
12
DLT itself appropriate.
13
is the dose-limiting toxicity that we are defining
14
the right way?
15
And we
We talked about MTD, but
Should we think about biologically effective
16
dose rather than thinking about DLT?
17
was also made clear that maybe we need to be
18
looking at minimum effective dose, what have you.
19
Anyway, probably we need to think about beyond MTD
20
and DLT, what are the other things that we can have
21
as our endpoints or the object that we want to
22
reach in these.
A Matter of Record (301) 890-4188
Of course, it
192
1
Are we using all the data that we have?
I'm
2
sort of partially saying no.
3
excellent examples in the last two days that we
4
have heard, where actually a lot of the data have
5
been used as well.
6
using as much of the data.
7
I think there were
So maybe sometimes we are not
But there could be also another reason that
8
we are not using all of the data.
We have to
9
understand that these phase 1 studies or
10
preclinical studies, they're data rich in some
11
sense, that on the same patient, you may have a lot
12
of measurements, but still they are a small number
13
of patients.
14
from the patients that enter into phase 3 trials.
15
In general, the phase 1 patients are more sick, and
16
therefore, you may not even observe the long-term
17
toxicity because they're no longer there.
18
don't have that experience even.
19
And these are patients very different
So you
How can we learn from past observations, and
20
how can we account for the limitations and
21
uncertainties?
22
there is a cost to it, for sure.
Can we be more efficient?
A Matter of Record (301) 890-4188
Yes, but
193
I think this was sort of brought up in some
1 2
of the presentations yesterday and today and best
3
practices.
4
have finished phase 1 and that's it.
5
preclinical data, it's done, but think of it as
6
more of a product life adaptive process, and we
7
have to keep improving based on this.
Our
There can be, of course, issues with this,
8 9
We have to stop thinking of this as we
and I will go through some of them in my later
10
slides.
11
finish these slides just before the session, so
12
bear with me.
13
that I was hearing in the last three sessions.
14
Let me see what I have.
I was trying to
I wanted to add as much as possible
We have approved many kinase inhibitors.
15
can review this data from phase 3 studies to
16
characterize what is an acceptable toxicity.
17
going back to the DLT definition.
18
have this accurately, so maybe we should go back
19
and look into what is it that we need to define
20
this; so redefine DLT.
21 22
We
I'm
Maybe we did not
Record when these toxicities actually occur, so what is the time to this toxicity, if it's a
A Matter of Record (301) 890-4188
194
1
delayed toxicity.
And estimate the exposure, how
2
long treatment was received before you saw such a
3
toxicity maybe.
4
observed dose modifications that were made, and use
5
this to go back into that product life cycle that
6
we had.
Statistical model based on these
So what can we further do?
7
Review data from
8
phase 2 studies as well; understand the limitations
9
and uncertainties in the PK/PD modeling.
It's not
10
going to be perfect.
I like that one of you said
11
yesterday that no model is a good model.
12
have to keep that in mind.
13
models have assumptions, so it is only as good as
14
the assumptions are true.
It's a model.
So we All
What is the missing piece of information
15 16
that we should have assessed in order to make a
17
better decision that we didn't have?
18
that?
19
looking.
20
are already doing that -- that was mentioned
21
yesterday -- and review data from phase 1 studies
22
also.
Can we assess
This is all retrospectively going back and And I think some of the companies here
What happened to patients beyond the first
A Matter of Record (301) 890-4188
195
1
cycle?
What dose modifications were made, and what
2
toxicities were observed beyond the first cycle? Do we have the right PD marker?
3
That's
4
another question that we need to ask as well.
5
immature data that were not used in the preclinical
6
or phase 1 studies that we could have used and that
7
would have led to better decisions. So these are questions that we have to go
8 9
Any
back and retrospectively look at the data and
10
answer them.
11
to say what can statisticians do here.
12
incredible amount of data, and these can be used
13
for priors in the statistical models.
14
simulate multiple, clinical scenarios.
15
of these examples yesterday.
16
critical to think beyond the 28-day cycle phenomena
17
that we have, and maybe use a new definition of
18
DLT.
19
Since I'm a statistician, I'm going We have an
We can We saw some
And I think it's
MTD is a question mark.
And I think it's
20
somewhat needed, or however you want to classify
21
it, MTD or BED or MED, what have you.
22
to have something in the product label as well.
A Matter of Record (301) 890-4188
But you need So
196
1
in that sense, you need to know what the patient
2
can tolerate.
3
I think the mistake that's being done is
4
when we are characterizing the efficacy, we always
5
give the confidence interval along with the
6
estimate.
7
summary that we are giving.
8
every patient will have exactly this survival just
9
because we put the median survival there.
10
We are saying that this is a population We are not saying
So similarly here, yes, we have an MTD, and
11
this is based on the aggregation of the data, but
12
there are always going to be some uncertainty about
13
it and some broad interval around it.
14
concept of variability is totaling the same in this
15
MTD determination.
16
I think the
We can also think of modeling toxicity and
17
efficacy, however we define this efficacy.
18
often, we are using only the response rate here.
19
But we are seeing that more and more, the phase 1-2
20
studies are being used for at least early
21
approvals, maybe accelerated approvals.
22
case, should we be optimizing both of them, and is
A Matter of Record (301) 890-4188
Most
So in that
197
1
there room for using both efficacy and toxicity in
2
these models?
3
rather than saying this is the fixed dose that
4
keeps going on and on, and use what you learn is
5
the key thing.
Model dose as a function of time
6
Back to this product life cycle, then, you
7
start, you go, and I think everybody does the best
8
that they can with what they have.
And you move
9
from in vitro to preclinical data.
Yes, probably
10
it's always going back retrospectively.
11
say I could have done some of other animal model or
12
phase 1 dose finding.
13
different.
14
point, you are making decisions based on what
15
information you have.
16
You can
I could have done something
But it's all in retrospect.
At that
So you come to phase 3, and then you see
17
that there are a lot of dose modifications.
18
think the sponsors have to clearly think about,
19
okay, now that I have this information, how can I
20
better inform the patients and physicians to give
21
this treatment to the patients?
22
So I think that's where you need to think about how
A Matter of Record (301) 890-4188
I
How can they dose?
198
1
can we loop around back again and what kind of
2
trials can we do?
3
and inform and do smaller clinical studies so that
4
we can inform better?
Can we do more of preclinical
I think we all came out saying that 3 plus 3
5 6
designs are not really efficient.
7
that it has no memory.
8
data.
9
you have to do it, okay.
10 11
It's just simply
It's not using all the
It's not the best of the designs.
But if
But you know that it's
not the best design. Model-based dose-finding trials are more
12
efficient for sure, but we should not be
13
blind-sided here, too, that, yes, these models also
14
assume things.
15
good as the priors that we use.
16
the Bayesian paradigm you have, you can learn from
17
what you're observing and then correct them as you
18
go along.
19
We use priors, and it's only as But at least in
So the feasibility question I thought was
20
discussed quite well.
And from what you have heard
21
from us -- at least from the FDA, those of us who
22
are here, and I think we can speak for the oncology
A Matter of Record (301) 890-4188
199
1
products -- that we are not saying you have to do
2
3 plus 3 for sure.
3
question some of the things that were raised was
4
that computing and statistical expertise are very
5
important, and it is expensive.
6
investigators and scientists is necessary.
7
are operational logistics in running some of these,
8
and there is certainly an upfront investment
9
necessary.
10
But then the feasibility
So there is a cost.
And buy-in from There
This is the cost to
11
doing that.
12
the same thing for every product.
13
loss to characterize the uncertainty in the
14
estimates, and that's what is lacking in what we
15
are doing right now.
16
But up front, you enlist, you can use It's certainly a
Use all the available data to make informed
17
trial changes, and we also discussed about the
18
therapeutic index versus MTD.
19
think it may be different for different products,
20
different diseases, but we need to have something.
21
Maybe even if you said this is the dose range that
22
we have to be treating, that's okay, but you need
A Matter of Record (301) 890-4188
Whatever be it, I
200
1
to have that. There are best practices that we'll share
2 3
here, and it was very informative; to improve
4
efficiency and use of preclinical models and phase
5
1 results to make go/no-go decisions.
6
a very nice way of using these data, not
7
necessarily to find the dose finding as well.
8
Preclinical models then to manage toxicity
9
post hoc.
I think it's
So you have already approved.
10
back.
11
concept as well.
You go
I thought that was a very interesting
Randomized phase 2 with 2 or more doses to
12 13
explore all of this.
I think we all agree, any day
14
we will take a randomized phase 2.
15
particularly all statisticians will say that we
16
want randomized studies.
17
about it, that if you come with a phase 2, 2 doses,
18
that's fantastic.
Believe me,
So there's no question
Continuous learning and improvement is
19 20
essential.
I think this process is there in
21
engineering, and I don't know why we don't use it
22
here.
Think of it as a product life cycle process.
A Matter of Record (301) 890-4188
201
1
Data and experience sharing that was done here I
2
think is very important among sponsors.
3
greatly improve efficiency.
4
each other.
5
patients will be benefit by this.
It would
You can learn from
And of course, more importantly, the
6
Thank you, and that's all I have to say.
7
(Applause.)
8
DR. KIM:
9
Up next, we are grateful to have
Dr. Lilli Petruzzelli from Novartis, who's going to
10
be discussing barriers to implementing integrative
11
and adaptive dose-finding trials.
12 13
Thank you.
Presentation - Lilli Petruzzelli DR. PETRUZZELLI:
Thank you for inviting me.
14
Thank you for staying the whole afternoon.
I am
15
going to be talking to you about the challenges of
16
implementing adaptive design trials, but I will be
17
focusing largely on the BLRM designs because that's
18
what we do.
19
already brought to the forefront in this meeting
20
already, so I don't think that -- I'll try to
21
consolidate a lot of what I've heard today and over
22
the last two days about the challenges in
I think most of the challenges you've
A Matter of Record (301) 890-4188
202
1 2
introducing the designs. This is really -- I think Stuart may have
3
summarized this.
4
his talk.
5
coming out of our -- into our first-in-human
6
studies.
7
and tolerability of the drugs.
8 9
I unfortunately couldn't be at
But we view this, sort of key challenges
We really want to understand the safety
It was really interesting to hear today how many people struggle with, really, what they do as
10
the MTD and a definition of the MTD.
11
understand what's happening as we move to the
12
higher range of these drugs, and largely because of
13
the PK variability.
14
excursions go outside of range, what's going to
15
happen to a patient?
16
Sometimes we miss those signals.
17
learn that from this first-in-human study.
18
We do like to
And we'd like to know if the
They're small numbers. But we do want to
Safety's first, and what we really try to do
19
is the optimal dose for our future development.
20
Many times we've used this adaptive design, and
21
we've also used the phase 2 study looking at two
22
different doses.
So this isn't the answer.
A Matter of Record (301) 890-4188
It
203
1
isn't like we always come up with one dose that we
2
want to move forward with, but we try to get as
3
much information to at least define what range is. So I think it is very reasonable to proposed
4 5
that you could come up with a dose range.
6
Sometimes you don't have a true MTD, and sometimes
7
you may want to look at two doses going forward.
8
We think that's a very reasonable model. One thing's for sure, we have definitely
9 10
moved away from the 3 plus 3.
11
BLRM in almost all of our studies.
12
largely because we are grouped -- the focus is on
13
novel molecules, the first-in-human studies in
14
combination and in single agents.
15
when we're combining with chemotherapy and standard
16
of care, where we may move away from the adaptive
17
design.
18
over 150 studies that we've used adaptive designs.
19
And in general, all of our combinations,
20
particularly of novel agents, have used that.
21 22
But do we use the In fact, we do,
There are times
But I would say over the years, it's been
I think no one needs to revisit the limits of 3 plus 3 design.
Where we've come to in this
A Matter of Record (301) 890-4188
204
1
model is that we spend a lot of time actually
2
working together early on, looking at historical
3
data, trial data, and then working very closely
4
with the statisticians. I intuitively am not a statistician, so I am
5 6
like wedded at the hip to my statisticians for
7
discussion and answering questions.
8
worked very closely with Stuart Bailey over the
9
last five years, and he knows.
And I've
For me, it's all
10
about the discussion and the collaboration that
11
really makes this work.
12
group.
13
forward is we rely very much on our relationship
14
with our investigators to make this work and their
15
input.
16
And so we do, we work as a
And what's really critical in moving
So what are the challenges to implementing
17
BLRM designs or an adaptive design?
You need the
18
expertise, and we have spent a lot of time building
19
that expertise.
20
general, there are turnovers.
21
core group always able to do this and always able
22
to learn?
Like any other industry in
A Matter of Record (301) 890-4188
So how do you make a
205
So that's a big focus of our implementation
1 2
of BLRM.
We have a lot of studies.
If we've
3
instituted 150 over the last decade, you can
4
imagine how many we have going at any one time,
5
where we're dose escalating.
6
simulations.
We need to run
We need to have input of data.
We have to have a core group of expertise.
7 8
But we also have to have education there because
9
new people -- you have new investigators.
We have
10
new clinicians coming in who've never seen the
11
BLRM.
12
can do everything, but this is new to them.
13
actually spend a lot of time educating people.
14
And they're smart.
They're complex.
They can write.
They So we
And the other is
15
communication.
16
both when we talk about how to implement the
17
designs, but also when we write about them.
18
that's a lesson learned I'll share with you.
19
again, it's collaboration.
20
amongst ourselves and also with our investigators.
21
And we spent a lot of time on that.
22
Speed.
Communication is a real challenge
And And
It's collaborating
They do have a potential to slow you
A Matter of Record (301) 890-4188
206
1
down.
2
patients, and they become costly.
3
patients to understand more.
4
We run the simulations.
5
there's no question that we have to balance speed
6
versus gaining information.
7
You start looking -- you're adding other We add more
We stop.
We pause.
They can be slower.
So
Building the model requires key input from
8
both the statistician and the clinician.
So early
9
on, as I said, the educational infrastructure is
10
critical, and we actually run courses.
11
group generously runs courses with lots of goodies
12
they give away, mostly candy, the whole year.
13
it's quite motivating, and people love them.
14
Our stats
And
They run different levels, so you can
15
actually feel like you're graduating in your
16
understanding of the BLRM.
17
can consistenly use it.
18
memory, and there's an experience, and there's been
19
changes that we've instituted as we've gotten more
20
familiar with them.
21 22
It's been fabulous.
We
So there's been really a
There has to be a lot of interplay between our preclinical tox people and our preclinical
A Matter of Record (301) 890-4188
207
1
scientists to understand the toxicology, proposed
2
starting dose, the predicted efficacious dose using
3
models, knowing all the weaknesses around those.
4
And we can't say enough, that we don't overstate
5
them, but we have to begin somewhere.
6
and getting really good and strong data really
7
matters.
So this data
8
Then the drug-drug interactions matter.
9
It's how fast you can go sometimes, what you're
10
worried about, your dosing intervals, what your
11
escalation intervals will look like.
12
factors in.
13
very important for interpreting the data.
14
matters?
15
lymphoid depletion that you're seeing in your tox
16
versus the liver?
17
things we worry about, and we give input on that.
18
That all
And then the input of the position is What
Are you going to really worry about the
I mean, there is a balance of
I wanted to take a break and take you
19
through -- and it won't be ceritinib, although I'm
20
quite familiar with ceritinib.
21
take you through another program where we've
22
implemented, to show you where we hit our -- what
A Matter of Record (301) 890-4188
But I wanted to
208
1
the challenges are to implementing it, and in
2
addition, how we use it.
3
in progress.
4
I would like to share with you how we think about
5
this.
6
This is a bit of a work
So I'm not sharing this publicly, but
This is a fairly typical set of preclinical
7
data that we gather together and we use.
There
8
might be a little bit more, but in this particular
9
drug, we knew it was a heme-specific [ph] drug.
We
10
knew we were going to go in a specific patient
11
population such as myeloma.
And we knew a couple
12
of features about the drug.
It had a potential to
13
have a time-dependent inhibition, DDI, but it was
14
very well tolerated in tox studies with basically
15
some aspirations seen largely because of the large
16
amount of the drug we had to give, and some
17
lymphoid cell depletions.
18
really worried about.
19
So nothing we were
We had a starting dose estimate initially of
20
150, but actually it turned out to be 70.
Because
21
of the potential for DDI, we ended up modifying the
22
protocol.
And we do feel that with all the work
A Matter of Record (301) 890-4188
209
1
that goes into doing these BLRMs, getting the
2
starting dose right from the beginning really saves
3
us a lot of time and avoids amendments.
4
a very simple dose escalation scheme.
5
So we had
So where do we run into our first hurdle?
6
And it really is all around the language.
We've
7
been doing this for 10 years, and what we've
8
learned is that consistent language in the
9
protocol.
We have tremendously talented physicians
10
in our group who write great protocols, and that we
11
were finding everybody was writing their own
12
version of this.
13
About five years ago, Stuart and I took all
14
the language together, and we became very
15
particular about it so there wasn't any confusion
16
when we went to sites, and when we went to the
17
health authorities.
18
consistent.
19
incorporated those changes, so we will modify our
20
standard template as we get questions.
21
we're good, but they still keep coming up.
22
So the questions became very
And over the years, we've actually
We think
Believe it or not, we've picked a core group
A Matter of Record (301) 890-4188
210
1
of investigators for our first-in-human we worked
2
with, so there's quite a bit of familiarity.
3
Sometimes when we go for health authority review,
4
we get a wider range of questions.
5
can actually defend and discuss, and some of them
6
actually prompt us to change the language.
7
protocol language consistency actually makes life a
8
lot easier.
9
simple fix, it's actually quite important.
10
Some of them we
So the
If I give people what looks like a
We get a consistent question depending upon
11
who reads it.
12
what are you going to do?"
13
us to take that data in and decide whether we're
14
going to stay at the dose, we're going to dose
15
de-escalate, and most of the time people want us to
16
stop and declare it the MTD.
17
to do quite a bit of education on this, to this
18
day, and I probably think we haven't gotten the
19
language perfect.
20
"Well, if you have 2 out of 6 DLTS, Well, our model allows
So we actually have
There is oftentimes a request for examples
21
on how the BLRM functions within a protocol.
22
avoided doing this because the protocols just get
A Matter of Record (301) 890-4188
We've
211
1
to be ridiculously lengthy, and sometimes the
2
examples don't always help.
3
things; we've seen lots of requests.
4
actually made a decision not to do those examples.
5
What really matters is an excellent
So that's one of the And we've
6
relationship with the sponsor, the investigator.
7
It's not just our internal communication, but it's
8
also our external and working with our
9
investigators.
People have to be on the telecon
10
calls for these decisions.
11
us what to do.
12
investigators and their impression of things that
13
really matter.
14
The model isn't telling
It's the input from the
What data is considered for dose escalation?
15
Well, we consider, obviously, the grade 3 DLTs, but
16
we also want to hear from the investigator about
17
the grade 2 AEs.
18
we have, we evaluate.
19
PK, the DLTs, and any other AEs.
20
We look at the PK, and if any PD Buy it really is around the
We limit our dose changes to a maximum of
21
dose doubling.
I will show you the example where
22
we actually had to go a little bit above because
A Matter of Record (301) 890-4188
212
1
our pill size, when we had to back off, didn't fit,
2
70 to 140.
3
cooperative and helpful for us in allowing us
4
slightly greater than doubling.
And the agency was really very
The model remember takes into account all of
5 6
the available data and can be run -- we can run the
7
model more than once to get the next dose level
8
that the model recommends.
9
recommendation.
But it's a
It's not what we do.
This is just
10
one point.
11
uses.
12
DLT rate is within the target dose interval.
13
These are the criteria that the model
It maximizes the probability that the true
The key that we always run into trouble with
14
is that it projects that the DLT rate -- that the
15
overdose, less than 25 percent probability that the
16
true DLT rate will exceed 0.33.
17
we always get into trouble with this model.
18
once we explain it and we have standard language,
19
people really feel comfortable with it.
20
And that's where But
The cohort sizes are specified, but we also
21
have flexibility around the cohort size.
22
begin to see toxicities or we want to learn more
A Matter of Record (301) 890-4188
So as we
213
1
about the PK, we can then add additional patients
2
and additional cohorts with a specified size, and
3
we do that quite often. So I wanted to show you a little bit about
4 5
the model because one way we always get questions
6
is around the provisional dose levels.
7
just what we're estimating, and usually it's dose
8
doubling.
9
and we're worried about accumulation, what happens
They're
But when we know more about the model
10
is we slow down sometimes at dose escalation, and
11
we'll fix that right from the beginning.
12
questions come up oftentimes of how we've chosen
13
our provisional dose levels, but they're only
14
provisional dose levels.
15
model recommends, and then the input from the
16
investigators, and the final decision resides in
17
the telecon, and that input is key.
So
And it really is what the
This is a case where we started at 70.
18
We
19
were able to go 150 based on our prescribed pill
20
side.
21
first dose level, to go 150.
22
investigators began to see grade 2 toxicity,
And the model said, after looking at the Then the
A Matter of Record (301) 890-4188
214
1
thrombocytopenia.
2
anxiety.
3
seeing early signs, and should we slow down.
4
there's a question.
5
escalate?
6
seeing any grade 2.
7
There was quite a bit of
Several patients had it and were we
Do we add more?
So
Do we dose
Well, the model said you're not really Go to 300.
Well, the investigators were quite nervous,
8
and they did a compromised dose, 250.
9
would have been fine -- 200 I mean.
I think 200 They could
10
have stayed at 150 as well.
11
decision not to escalate as quickly based on
12
clinical data that was coming in.
13
grade 3 thrombocytopenia, and people said, are we
14
really seeing early signals of toxicity?
15
But that was a
Then they hit a
Now, this was in a heme population.
It was
16
easy to see these.
They're heavily pretreated.
17
was easy to pick up these hematologic toxicities.
18
But what happen was -- and you'll see there was
19
continued dose escalation and why.
20
a little bit in this process, and we watched the
21
patients over several cycles.
22
yes, they had an initial drop in cycle 1, but
A Matter of Record (301) 890-4188
It
We slowed down
And lo and behold,
215
1
everybody leveled out.
2
manageable thrombocytopenia that initially, if we
3
had just reacted to the first cycle, we might have
4
stopped at 150.
5
So there was actually a
That may well be the right dose.
We're not
6
certain yet.
I mean, at that point we weren't
7
certain.
8
that was limiting dose escalation.
9
what we learned was when we started to get to 500,
But certainly, this was the only thing And in fact,
10
we saw a more dramatic and consistent decrease in
11
the platelet count, and that became more worrisome.
12
So we felt that there was eventually going
13
to be dose where thrombocytopenia became a problem,
14
but there was a lot more room to move, and it was
15
only watching the chronic behavior that we actually
16
became a little bolder with the algorithm, not more
17
than conservative.
So it can work in both ways.
18
So we ended up finishing dose escalation.
19
We actually ended up going all the way up to 700.
20
We called the highest tolerated dose 500
21
milligrams.
22
patients in the end in the study because the PK was
We had dosed about 50 some odd
A Matter of Record (301) 890-4188
216
1
quite variable, and we added some extra cohorts to
2
understand better about the PK profile.
3
came up with a recommended dose of 300, and I'll
4
explain that a little bit in a second.
5
So additional cohorts.
And we
The advantages of
6
this -- or the additional cohorts at lower doses
7
enabled us to understand the thrombocytopenia
8
better, enabled us to feel more secure about the
9
plateau.
Only small numbers -- so we got better
10
sense of the safety and tolerability by being able
11
to add additional cohorts.
12
to then use this extra data to try to model both
13
the toxicity cutoff, and as well as efficacy.
And then we were able
Here, what we saw were 10 total DLTs, and
14 15
that makes everybody a little bit anxious.
16
the standard grading of grade 3 with bleeding or
17
grade 4.
18
really only saw thrombocytopenia and some
19
neutropenia.
20
it was only 10 out of 53.
21
level.
22
doses that made patients uncomfortable with this
It was in a heme population.
We used
And we
Fifty-three patients were dosed, so It was an acceptable
And really, it was fatigue at the higher
A Matter of Record (301) 890-4188
217
1
drug.
And they really wanted to stop and dose
2
reduce when they got to 500 because of the severe
3
fatigue, almost not wanting to get out of bed.
4
it was a tolerability issue.
So
What we said, well, how are we going to pick
5 6
the dose?
And this is a case where we might not.
7
They were in heme.
8
will go in.
9
do was take the preclinical dose and then take the
This is likely the indications
But in general, what we were able to
10
responses.
And it was an interesting drug because
11
most patients have stable disease or a modest
12
response after we hit 150 milligrams.
13
was -- above 200 was a real cutoff where the
14
majority of patients had some degree of
15
stabilization of their disease or response.
16
felt fairly comfortable that if we're above 200,
17
we're likely going to be in an efficacious range.
So it really
So we
Could we go higher if we found diseases
18 19
where we wanted to potentially push us higher; they
20
were dependent on this target?
21
enough about this drug.
22
safely.
We feel we know
Could we go to 400?
Yes,
We picked 300 because it gave us some dose
A Matter of Record (301) 890-4188
218
1
reduction room, but it's fairly arbitrary.
2
could argue when we go to phase 2 that we look at
3
more than one dose, and that is an option.
4
And one
This is case where getting more data we feel
5
helped us understand the molecule better.
It
6
actually enabled us to dose escalate more rather
7
than stop sooner, and that's for us the power of
8
the model.
9
it's been ongoing for us -- is the
Where we find the challenges -- and
10
flexibility -- is the language.
11
to do also is to use this system more to have more
12
flexibility around looking at different doses and
13
different schedules, and different dosing
14
intervals.
15
And what we'd like
We feel very strongly -- and I think we
16
heard this from everybody.
The MTD is really not
17
all what this is about.
18
and tolerability and picking the phase 2 dose is
19
really critical.
20
dose escalation to factor in key data is a
21
challenge.
22
We sometimes predefine that.
Understanding the safety
Defining the requirements for
Include the data from late toxicities.
A Matter of Record (301) 890-4188
219
There are some molecules where we know we're
1 2
going to see late toxicities, and we will predefine
3
the longer period.
4
and that data comes in later, and we don't always
5
have that captured.
6
within this to inform the model and make decisions
7
with the clinicians around the late toxicities.
But we do have the flexibility
The continued education about the method and
8 9
But sometimes we're surprised,
the sharing best practices is key, and it remains a
10
challenge as new people come into the industry and
11
as new people do reviews.
12
that we're very willing to -- we've made real
13
concerted effort to be available to help people
14
with that, but I think to implement this, that
15
infrastructure is key.
And that's something
The real key is also, this isn't always the
16 17
fastest way to run the study.
It can sometimes be
18
slower.
19
MTD and dose for expansion in eight months with a
20
terrific number of patients and everything.
21
move very quickly.
22
we have new molecular entities?
We do have studies where we achieved our
We can
Will we run into toxicity when
A Matter of Record (301) 890-4188
We sometimes have
220
1
to go slower.
2
expensive to run these studies as we add additional
3
patients.
4
And it can become much more
So those are the challenges that we've seen
5
facing us as we move this forward.
6
to use it, and becoming more consistent with it has
7
really made it easier for us.
8
(Applause.)
9
DR. KIM:
But we continue
So thank you.
I would like to declare our
10
workshop a success because we are very efficient,
11
and we are way ahead of time here.
12
that I think a lot of people have afternoon flights
13
and want to get out of here quicker.
14
could take a 20-minute break instead of the
15
allotted 30-minute break, and come back here at 10
16
after 2.
17
practical considerations of implementing dose
18
optimization strategies.
19
time around.
I'm mindful
so maybe we
And after that, Dr. Shaw will talk about
So a 20-minute break this
Thank you.
20
(Where, at 1:50 p.m., a recess was taken.)
21
DR. KIM:
22
I hope everybody had a nice break.
Next, it's my honor to introduce Dr. Alice Shaw
A Matter of Record (301) 890-4188
221
1
from the Massachusetts General Hospital, to
2
introduce some of the practical considerations of
3
implementing dose optimization strategies in the
4
clinic.
5
clinician --
Dr. Shaw is an actual real-life
6
(Laughter.)
7
DR. KIM:
8 9 10
-- and is going to give us some
real-world experience with these types of trials. Presentation - Alice Shaw DR. SHAW:
Thank you, Geoff.
And thanks to
11
the organizers for inviting me to speak here today.
12
And as Geoff said, I will be sharing with you the
13
clinical perspectives on some of the dose
14
escalation designs we've talked about and
15
strategies for dose optimization.
16
We've been talking about a number of
17
different tyrosine kinase inhibitors, and I thought
18
the best way for me to probably frame my thoughts
19
on these questions of dose optimization was in the
20
context of some specific tyrosine kinase
21
inhibitors, and specifically ALK inhibitors.
22
So just to remind you, we already heard
A Matter of Record (301) 890-4188
222
1
about a number of them, and we heard some
2
background as well.
3
rearrangement defines a very specific subset of
4
lung cancer patients.
5
patients, roughly 5 percent, but it is a
6
significant number.
7
head, 5 percent does translate into a large number
8
of patients because there are over 1.6 million new
9
cases of lung cancer worldwide every year.
Just briefly, we know that ALK
It's a small percentage of
If you do the math in your
10
5 percent actually ends up working out to a
11
relatively large number of patients.
12
So
The important things about these patients,
13
we all have heard the story.
14
have some unique clinical pathologic features, but
15
importantly, these patients are really responsive
16
to small molecule ALK inhibitors, and not just one,
17
but actually can be responsive to multiple ALK
18
inhibitors given sequentially.
19
These patients do
This has of course now spurred the
20
development of numerous ALK inhibitors.
21
Geoff, I would say maybe it's not the ceritinib
22
era; maybe it's the ALK era, because there are
A Matter of Record (301) 890-4188
And,
223
1
numerous ALK inhibitors now in clinical
2
development.
3
We're all familiar with crizotinib, which I will
4
spend some time talking about.
5
next-generation ALK inhibitors are shown here.
6
These in general are more potent than crizotinib
7
and have been designed to overcome crizotinib
8
resistance.
They're 9 of them as shown here.
But other
There are a number of them.
9
There are
10
different stages of development, as you can see
11
here.
12
three most advanced ALK inhibitors in the clinic,
13
crizotinib, ceritinib and alectinib, and now the
14
development of these drugs, particularly the
15
development and the dose selections have gone from
16
a clinical perspective.
And what I was going to focus on is the
Let's start with crizotinib, which as we all
17 18
know is the first ALK inhibitor tested in the
19
clinic.
20
actually originally developed by Pfizer to be a
21
very potent c-MET inhibitor and was only
22
subsequently found to inhibit a number of other
And as we heard yesterday, crizotinib was
A Matter of Record (301) 890-4188
224
1
tyrosine kinases such as ALK and the related
2
receptor tyrosine kinase, ROS1.
3
Of course, now the story is somewhat old
4
because over the last seven years now, there have
5
been multiple studies of crizotinib that have
6
established both its efficacy and safety in
7
treating patients who have advanced ALK rearranged
8
lung cancer.
9
it summarizes all the clinical data that we have on
And I like showing this slide because
10
crizotinib from the phase 1 study, which we're
11
going to focus on, to the phase 2, and both of the
12
phase 3 studies.
13
I think it's been remarkable how consistent
14
the efficacy signal has been through all of these
15
studies in terms of the response rates and also the
16
durations of response.
17
remarkable is that the safety signals that were
18
first identified and investigated in the phase 1
19
trials have also been very consistent through the
20
larger confirmatory studies that have been
21
performed.
22
And also what I have found
So let's talk about the crizotinib phase 1
A Matter of Record (301) 890-4188
225
1
study, and this is actually -- this was a
2
traditional 3 plus 3 study design, which of course
3
we've all just talked about the last two days not
4
being such a good study design.
5
study design was very successful in identifying, I
6
think, the best dose of crizotinib that we're using
7
in the clinic right now.
8 9
But in fact, this
This is the schematic of the study design that was used in the phase 1 crizotinib study;
10
started at a low dose of 50 milligrams once a day
11
and went as high as 300 milligrams twice a day.
12
This study started back in 2006, summer of
13
2006, so quite a few years ago, before ALK was even
14
known to be a target in lung cancer.
15
original intention really was to do the dose
16
escalation, and then to have expansion cohorts
17
particularly focusing on patients with MET
18
abnormalities, since, as I mentioned, it was really
19
known to be a potent MET inhibitor.
20
And the
As we heard yesterday, there was a fair
21
amount of good timing and I would say serendipity
22
in that this study was well on its way.
A Matter of Record (301) 890-4188
Dose
226
1
escalations were going well.
2
discovered in 2007 by Dr. Mano and grew up to be a
3
target in lung cancer.
4
study was already well into the therapeutic range
5
of dosing.
6
When ALK was
And at that time, this
The asterisks here indicate where the first
7
ALK-positive patients enrolled.
So these are
8
patients who were screened for ALK rearrangement.
9
As I said, the study was open and enrolling, so
10
those patients, the very first one, in fact,
11
enrolled in the highest dose cohort, the 300-
12
milligram, twice-a-day cohort.
13
the perfect setup where you have a new target, and
14
you happen to have a targeted therapy already ready
15
to go and in the therapeutic range.
16
ALK-positive patient then went in to a slightly
17
lower cohort but still in the therapeutic range.
18
So already you had
The next
As we were going through this study, in the
19
cohort 3, we did have one DLT, a grade 3 ALT
20
elevation, and then the next DLTs were seen in
21
cohort 5, the 300-milligram, twice-a-day cohort.
22
These are two grade 3 fatigues that counted as
A Matter of Record (301) 890-4188
227
1 2
DLTs. I should note that that first ALK-positive
3
patient that I mentioned who went into the study at
4
this high dose, actually he did not have a DLT, but
5
he did experience a grade 3 ALT elevation.
6
happened after the 28-day DLT period, so he did not
7
count as a DLT.
8
and actually, unfortunately, couldn't tolerate the
9
great challenge with crizotinib.
10
That
But this patient did have to hold,
Because of these 2 DLTs, of course in the
11
3 plus 3 design, the next dose was going to be
12
reduced, and this was a bit of discussion between
13
the study teams at Pfizer and the clinical
14
investigators of what the next dose down should be.
15
Should it be the 200-milligram, twice-a-day dose or
16
should it be an intermediate dose?
17
I think there were really two main reasons
18
driving the interest to look at the 250-milligram,
19
twice-a-dose, which of course is our standard dose.
20
One was that the 200-milligram BID dose had been
21
extremely well tolerated in these 7 patients.
22
There really were almost no AEs at all in these 7
A Matter of Record (301) 890-4188
228
1
patients, so that seemed like a very easy dose for
2
patients to tolerate.
3
had some emerging PK data at the time, suggesting
4
that the 200 BID dose would allow patients to reach
5
a dose that was very close to the level needed for
6
effective ALK target inhibition.
7
The other issue is that we
This shows some more mature PK data, showing
8
you where the predicted level for target ALK
9
inhibition would be, and the green here indicates
10
where the 200-milligram dose, where it lies
11
relative to this.
12
So there was interest in looking at a dose
13
maybe slightly higher than 200, and hence, the 250-
14
milligram BID dose was explored.
15
DLTs in that dose, and that became the dose that
16
became the standard of that dose going forward for
17
the expansions and the phase 2.
18
There were no
In the PK analyses, you can see that the
19
half-life for crizotinib, as we also have heard, is
20
on the long side.
21
over the years, many times it's come why is this
22
drug dosed twice a day instead of once a day, which
It's over about 53 hours, so
A Matter of Record (301) 890-4188
229
1
would have made a lot more sense.
And again, a lot
2
of that was driven by what we were seeing in the
3
clinic. In the once-a-day dose cohorts, we had seen
4 5
a fair amount of nausea and vomiting occurring
6
several hours after dosing, and that seemed to
7
coincide with when the Tmax was.
8
that by doing BID split dosing, that we might
9
decrease the Cmax and actually improve the nausea
So the hope was
10
and vomiting.
And that seemed to actually help.
11
So that's the explanation for the BID dosing.
12
as you can see, there is linearity in terms of the
13
PKs.
And
14
The last point I'll make about this is that
15
the phase 1 crizotinib study did have a food study,
16
a food substudy, and that showed that there was no
17
significant food effect on the crizotinib PK.
18
Here are the adverse events that were noted
19
in the phase 1 study.
This is the whole phase 1
20
study now, including the dose expansion.
21
consistent with what we saw early on in the dose
22
escalation, this drug was very well tolerated in
A Matter of Record (301) 890-4188
And
230
1
those patients treated at the standard 250 BID
2
dose.
3
effects yesterday.
4
mild it's seen, and most patients who are treated
5
with crizotinib, we've heard about the GI side
6
effects.
We've again talked about some of these side The visual disturbance is very
What's interesting about crizotinib is that
7 8
although we see these GI side effects, they really
9
don't appear to worsen over time the way we have
10
seen with ceritinib, which we'll get back to.
I
11
would say the only cumulative toxicity that we've
12
really noted with crizotinib in the clinic is
13
edema.
14
predict in preclinical testing.
15
this edema, even though it builds over time in
16
patients who have been on one, two, three years,
17
it's something that can be, in general, pretty well
18
managed.
That is one that is a little bit harder to But fortunately,
In terms of more severe toxicities with
19 20
crizotinib, fortunately, there really weren't that
21
many.
22
saw in dose escalation.
And again, this is consistent with what we There are some grade 3/4
A Matter of Record (301) 890-4188
231
1
elevations in liver enzymes.
We did see some
2
fatigue at the 250 BID dose and hypophosphatemia,
3
which is a class effect of ALK inhibitors, and then
4
of course some pneumanitis, which we see rarely
5
with many of the tyrosine kinase inhibitors.
6
overall, I think the toxicity profile that we saw
7
and the dose expansion was very consistent with
8
what we had predicted we would see based on the
9
results of the dose escalation.
But
Just to quickly summarize about crizotinib
10 11
and how the dose finding worked from a clinical
12
perspective, I would say in this particular case,
13
in this case of crizotinib with ALK, the
14
traditional 3 plus 3 dose escalation design was
15
successful.
16
a very good dose for our patients.
I think 250 milligrams twice a day is
Several of these confirmatory studies, which
17 18
I showed you, the phase 2 and, importantly, both of
19
the phase 3 studies of crizotinib have shown
20
similar efficacy and safety.
21
really, crizotinib is very well tolerated in the
22
clinic.
And as I said,
We only occasionally have to dose reduce
A Matter of Record (301) 890-4188
232
1
crizotinib in the clinic, and that's usually for
2
side effects such as elevated liver enzymes and
3
sometimes rarely fatigue. Finally, I'll just make a note that not all
4 5
of these side effects of crizotinib, but many of
6
them were predicted by preclinical toxicology
7
studies.
8
original consent form for patients who are
9
enrolling in the phase 1 portion of the crizotinib
I went back last night and looked at our
10
study.
11
already noted for crizotinib are already listed
12
there as being potential side effects based on the
13
toxicology studies.
14
And many of the side effects that we've
So let's now move to ceritinib, which I know
15
we've heard a lot about.
16
that much detail since we've already heard so much
17
about it, but I just wanted to share, again, the
18
clinical perspective on how this drug was developed
19
and how we ended up at the current standard dosing,
20
which is 750 milligrams once a day, under fasting
21
conditions.
22
I'm not going to go into
So again, we already saw that ceritinib is
A Matter of Record (301) 890-4188
233
1
very potent against ALK, also somewhat more
2
selective than crizotinib.
3
anti-MET activity.
4
crizotinib.
5
that supported the activity of ceritinib against
6
various models of ALK-driven cancers.
7
importantly, we had generated a lot of data looking
8
at ceritinib in terms of overcoming crizotinib
9
resistance, specifically crizotinib resistant
10 11
It doesn't have any
It's structurally distinct from
There was a lot of preclinical data
And I think
mutations. This is kind of a complicated slide, but
12
it's looking at BAF3 models that are expressing
13
different EML4-ALK fusions that have different
14
crizotinib resistance mutations as shown here and
15
comparing the efficacy or activity of crizotinib
16
versus ceritinib, and inhibiting each of these
17
different mutant forms of ALK.
18
that for the red bars, which is ceritinib, it's
19
quite a bit more active, and crizotinib clearly can
20
overcome a lot of the crizotinib resistance
21
mutations.
22
And you can see
I'm indicating these by red arrows.
On the other hand, some mutations, even the
A Matter of Record (301) 890-4188
234
1
gatekeeper mutation, the L1196M, even though
2
ceritinib does have activity, you can see that it's
3
not as active against the gatekeeper mutant form as
4
it is against the wild type ALK.
5
talked about the decision to kind of err on the
6
higher side of ceritinib, and a lot of that was
7
motivated by our desire to be able to inhibit some
8
of the common crizotinib resistance mutations that
9
we were seeing in the clinic.
We've already
10
Again, this was already discussed very
11
nicely by Dan earlier today, but just again to
12
remind you, here are the different dose escalation
13
cohorts that we used of ceritinib, ranging from
14
50 milligrams once a day to 750 milligrams once a
15
day,
16
enrolled into each of these cohorts, a total of 59
17
patients, 54 of whom are evaluable for DLTs.
18
again, we've already heard a lot about BLRM that
19
was used to guide our dose escalation decisions.
20
As we already discussed, patients were
And
I just wanted to make two other points about
21
this study, again, from a clinical side of things.
22
This study did allow us to start cohorts with one
A Matter of Record (301) 890-4188
235
1
patient in each cohort because there was a
2
concern -- well, we really wanted to minimize
3
treating patients at doses that were known to be
4
subtherapeutic.
5
have single patient cohorts, although we enrolled
6
actually 2 in the lower cohorts.
7
So the protocol did allow us to
The protocol also did allow us to backfill
8
cohorts at doses that were already cleared for
9
safety, so that's why you can see some of these
10
cohorts became quite large.
11
very useful information that we were able to gather
12
in terms of both efficacy and safety.
13
And I think that was
So again, I wanted to put this in the
14
context of the clinical situation that was
15
happening at this time.
This study was ongoing
16
starting in early 2001.
And as many of you know,
17
crizotinib was not yet approved.
18
approved in August of 2011, and there were many,
19
many patients, actually, who had been identified as
20
ALK positive who were in crizotinib and now were
21
failing crizotinib.
22
ceritinib was the only ALK inhibitor, only
It was actually
And at this time, in 2011,
A Matter of Record (301) 890-4188
236
1
next-generation ALK inhibitor out there.
So there
2
many, many patients who were trying to get on this
3
study, and we were actually seeing early on very
4
promising signs of activity. This is one of the patients who had failed
5 6
crizotinib before.
She enrolled in one of the
7
early cohorts, 400 milligram a day, and she had
8
this beautiful PR.
9
lower doses like the 300-milligram dose also having
We had patients enrolling at
10
these nice responses.
And some crizotinib naive
11
patients outside this country also went on the
12
study at even lower doses, like 200 milligrams, and
13
we saw responses. So there was already a lot of excitement
14 15
about this study from a clinical investigator
16
standpoint, and even patients were aware that there
17
was another option, potentially, after crizotinib.
18
So I think there was this feeling of urgency and
19
trying to speed the development of this drug in the
20
clinic.
21 22
Here are the DLTs that we observed in this ceritinib phase 1 study.
Our first DLTs occurred
A Matter of Record (301) 890-4188
237
1
at the 400-milligram dose where we had a grade 3
2
ALT elevation, and we also had a grade 3
3
hypophosphatemia, which we counted as a DLT,
4
although, as I said earlier, this is a known class
5
effect of ALK inhibitors.
6
We then had 2 grade 3 DLTs at the
7
600-milligram dose, including diarrhea and
8
dehydration, although I note that this patient
9
actually had been off of the drug when they
10
developed the dehydration, and then two more DLTs
11
at the 750-milligram dose.
12
GI, grade 3 diarrhea, as well as grade 3 vomiting,
13
and then another patient who had grade 2 diarrhea
14
that was intolerable, as well as grade 1 nausea.
15
Fortunately, in all cases, these toxicities did
16
resolve with dose holding.
17
one of these patients was able to resume study drug
18
at a lower dose.
19
Again, one of them was
Actually, I think every
At the time we had that information, we were
20
also generating the information on PKs.
21
this is just to show you here is what we predicted
22
to be the minimal level needed for inhibition of
A Matter of Record (301) 890-4188
And again,
238
1
ALK.
2
wild type ALK.
3
we knew some of the patients enrolling on this
4
study were crizotinib resistant because they had
5
resistance mutations, so those mutations might in
6
fact increase or require a higher dose of
7
ceritinib.
8
level.
9
Actually, this is the level of inhibition for So, again, our issue here was that
So we knew we had to at least hit this
So here's the 200-milligram dose, and here's
10
how high we went with the 750-milligram dose, and
11
there was a linear relationship here.
12
We saw this slide before.
How did we end up
13
at this current dose of 750 milligrams for the MTD
14
of ceritinib?
15
very nicely, so I'm not going to go into too much
16
detail about this.
17
into account all of the data, did actually allow us
18
to consider opening a 900-milligram dose cohort.
19
Well, again, Dan went through this
But the BLRM actually, taking
In fact, I very vividly remember this
20
telecon where we were talking with the Novartis
21
study team and all the clinical investigators about
22
the toxicities that we were seeing along the way,
A Matter of Record (301) 890-4188
239
1
including the DLTs, particularly at the 750 dose,
2
and could we actually imagine enrolling patients to
3
a higher dose, which the BLRM was saying might be
4
okay, and nobody could.
5
be something that was not going to be clinically
6
feasible for our patients.
So this really was felt to
We also did talk about how many
7 8
patients -- the nausea and the diarrhea with
9
ceritinib do not happen necessarily right away.
So
10
again, in this study the DLTs were primarily
11
examined in the DLT period, which was the first
12
cycle or 21 days.
13
the nausea and diarrhea could accumulate and could
14
worsen, even past that 21-day period.
And what we were seeing is that
We had a number of patients who were having
15 16
more significant GI side effects past cycle 1.
17
They weren't being counted as DLTs.
18
we had to hold their dose and dose reduce them.
19
because of that, there was a lot of concern about
20
going anywhere near higher than 750 milligrams a
21
day.
22
Nevertheless,
Then of course, we were talking about the
A Matter of Record (301) 890-4188
So
240
1
potential benefits of going to higher doses in
2
terms of overcoming resistance mutations, in terms
3
of penetrating well into the CNS because CNS
4
relapses are so common in these patients.
5
in the end, everybody was in agreement that we
6
would take forward the 750-milligram dose.
7
And so
However, there was sort of an agreement that
8
we would be looking at the first 10 patients who
9
enrolled at the 750-milligram dose in dose
10
expansion to make sure that those patients really
11
could tolerate the medicine.
12
were no DLTs in this group of patients, but again,
13
just in cycle 1.
14
And in fact, there
Here we have some data.
This is data that I
15
presented at ESMO back in 2012, looking at all of
16
the patients in dose escalation, as well as some of
17
the patients who are now enrolling into dose
18
expansion at the 750-milligram cohort.
19
can see, GI side effects really predominate here:
20
nausea, vomiting, diarrhea, elevated liver enzymes.
21 22
And as you
I think, as Dan was saying earlier, it seems when you look at the table, it's like there's no
A Matter of Record (301) 890-4188
241
1
definite dose dependency to these side effects.
2
But I would say these tables and the way we as
3
clinicians grade the toxicities, it's very hard to
4
sometimes capture these differences.
5
say, absolutely, there is a dose dependency, and
6
patients do much, much better in terms of side
7
effects at lower doses like the 400- or
8
500-milligram dose compared to the 750-milligram
9
dose.
10
And I would
Just to summarize the ceritinib story and
11
how we ended up settling on the 750-milligram dose,
12
we used the BLRM to help us decide on the MTD.
13
as I said earlier, and I wanted to emphasize, these
14
decisions really were based primarily on the
15
incidence of the DLTs that we were seeing in
16
cycle 1, the first 21 days of treatments.
17
think there was this interest, both on the part of
18
the sponsor and the investigators, to choose a
19
slightly higher dose because we really wanted to
20
maximize efficacy in these patients who were
21
failing on crizotinib, particularly in the CNS.
22
But
And I
I would say that in practice, now that the
A Matter of Record (301) 890-4188
242
1
drug has been approved for over a year, and we've
2
been using the drug both on study still, but also
3
off study commercially, that ceritinib dosing at
4
the MTD, under fasting conditions, which is how we
5
performed the study, really is very difficult for
6
most patients to tolerate.
7
necessarily happen right away.
8
through a month or 2 months of dosing, but in my
9
experience, every single patient I have ever
And again, this doesn't Patients might go
10
treated has needed to have a dose reduction, except
11
for one patient.
12
high dose, and lower doses absolutely are better
13
tolerated.
14
So I do think that this is a very
As we heard earlier today, there is a
15
significant food effect with ceritinib on PK and I
16
would say AEs.
17
use ceritinib who was out in New Zealand, and this
18
was an ALK-positive patient who had failed on
19
crizotinib.
20
for her to try a new drug.
21
treatment options, so she went on to ceritinib, and
22
she had a really tough time with the side effects.
We had a patient on compassionate
And really, this is last-ditch effort She had no other
A Matter of Record (301) 890-4188
243
She was on her own supply of this drug, so
1 2
she started experimenting with different ways of
3
doing the drug; of course, had to bring down the
4
dose, tried split dosing.
5
found was that taking the medicine with food made a
6
big difference.
7
email, and in the title, the headline was, "Stop
8
Torturing Your Patients:
9
she really felt like this offset a lot of the side
But really what she
In fact, I remember she sent me an
Allow Them to Eat."
And
10
effects, particularly the nausea and vomiting and
11
the upper GI discomfort that a lot of patients
12
have.
13
So again, as we heard, there is a global
14
food effect study, that is now underway, of
15
ceritinib, and I really think that this is going to
16
be the key to finding the optimal dose of ceritinib
17
that our patients can tolerate and that will be
18
effective for them as well.
19
In just the last seven minutes, I wanted to
20
talk about a third ALK inhibitor, and this is
21
alectinib, which, fortunately, we haven't talked
22
about yet.
And alectinib was originally developed
A Matter of Record (301) 890-4188
244
1
by Chugai, and now it's being developed by Roche.
2
It's already approved in Japan and does have FDA
3
breakthrough therapy designation.
4
Like ceritinib, alectinib is another
5
next-generation ALK inhibitor, chemically a very
6
different structure than either crizotinib or
7
ceritinib.
8
that alectinib is very potent against ALK, also
9
very potent against a number of the different
Also, lots of preclinical data showing
10
crizotinib resistance mutations, including the
11
gatekeeper mutation, lots of in vivo data showing
12
that alectinib is active against different
13
ALK-driven lung cancer models.
14
Actually, there are already studies going on
15
in Japan, but in the U.S., this led to a phase 1/2
16
study of alectinib, specifically in ALK-positive
17
patients who had already received crizotinib.
18
this is the study design here.
19
modified 3 plus 3 study design.
20
to the 3 plus 3 design.
And
This is actually a Again, we're back
21
Here, the way the study was set up is that
22
actually 3 patients would enroll into the cohorts,
A Matter of Record (301) 890-4188
245
1
and then those 3 patients would be followed
2
throughout the DLT period.
3
then that cohort would enroll up to 7 patients
4
total.
5
enable better PK, more robust and better PK data
6
from a slightly larger sample size.
7
that cohort cleared, the same thing was repeated in
8
the next higher cohort.
9
If there were no DLTs,
And the rationale for that was just to
And then once
This study went from 300 milligrams,
10
twice-a-day dosing, up to 900 milligrams twice a
11
day in the dose escalation.
12
cohorts that were done at the higher doses of 600
13
and 900 milligrams twice a day.
There are now bridging
14
In the initial part of the study, the 300-
15
to 900-milligram dose escalation phase, there were
16
actually no DLTs, and alectinib seemed to be
17
extremely well tolerated.
18
these doses being chosen to be the doses for the
19
two bridging cohorts.
20
And that's what led to
Now, in these cohorts, the 600-milligram
21
twice a day was very well tolerated, no DLTs again.
22
But the 900-milligram, twice-a-day cohort, there
A Matter of Record (301) 890-4188
246
1
were two DLTs in 6 patients.
One was a grade 3
2
headache, and one of my patients who had actually
3
untreated but asymptomatic brain mets when she
4
enrolled.
5
started on alectinib.
6
as a DLT.
7
had grade 3 neutropenia that did not resolve after
8
7 days.
And she developed a headache when she This was severe and counted
And then there was another patient who
So there were these 2 DLTs here. I should note the other toxicities that were
9 10
seen besides the two that I mentioned were fluid
11
retention.
12
grade 2, and that did respond to dose reduction,
13
also a grade 2 fatigue 760-milligram dose, and not
14
a DLT but that also responded once the patient was
15
dose reduced to 600.
16
other toxicities that were not attributed to the
17
drug.
18
That was seen at the higher dose but
And then there were a few
Like ceritinib, early on there were signs of
19
activity of alectinib in patients who had
20
previously failed crizotinib.
21
waterfall plot duration on study shown down here.
22
And down here, you can see these cohorts of
A Matter of Record (301) 890-4188
So this is a
247
1
patients that enrolled, and the dose is ranging
2
from 300 to 900.
3
right?
4
was designed.
5
is a very active drug, and this was what we knew as
6
the study was ongoing.
These are small cohorts, though,
Most of them are 7.
That's how the study
But it gives you a sense that this
7
Here are some more of the common side
8
effects that were seen with alectinib in this
9
study.
We've already talked about some of these.
10
Some of these are expected from ALK inhibitors.
11
would say the unique ones with alectinib include
12
myalgias and increased CK, as well as the
13
photosensitivity.
14
aren't surprised to see with an ALK inhibitor.
15
I
But many of the other ones we
In the end, the dose that was chosen, not as
16
the MTD -- the MTD was not defined, actually, but
17
the dose that was chosen as the recommended phase 2
18
dose was 600 milligrams twice a day.
19
the rationale for this, and this was actually
20
outlined.
21
but this is outlined very nicely in Dr. Gadgeel's
22
paper in Lancet Oncology.
And here's
I remember this on the telecon as well,
A Matter of Record (301) 890-4188
248
So really, there were no DLTs at all up to
1 2
the 760-milligram dose cohort.
And then there were
3
those 2 DLTs at the 900-milligram, twice-a-day
4
cohort, and that was the bridging cohort, although
5
one of them was the headache in the patient with
6
non-brain metastases. As we saw in those waterfall plots, we saw
7 8
really nice activity, even at some of the lower
9
doses.
But in the 600-milligram cohort, we saw a
10
pretty high response rate.
I think it was in the
11
60 to 70 percent range, and there wasn't a clear
12
increase in the response rate by going up to the
13
760 or the 900-milligram doses, although again,
14
it's really small cohorts. The PK data wasn't that clearly shown, but I
15 16
have since heard from Roche colleagues that the PK
17
is linear between 300 and 900 milligrams twice a
18
day.
19
choosing a dose that seemed to be very active,
20
which the 600-milligram dose was, and also one that
21
seemed to have basically almost very few side
22
effects.
However, I think there was a real interest in
And hence, 600 milligrams twice a day was
A Matter of Record (301) 890-4188
249
1 2
declared as the recommended phase 2 dose. Now, my question really has always been, is
3
this really the correct dose?
4
kind of the opposite situation to what we've seen
5
with ceritinib.
6
dose potentially to low for our patients?
7
In fact, this is
I guess my question is, is this
This is a situation that we see,
8
unfortunately, very commonly, which is CNS relapse
9
in ALK-positive patients when they've been treated
10
with ALK inhibitors, either crizotinib by itself,
11
or crizotinib and ceritinib, or other ALK
12
inhibitors.
13
is very active in the CNS, has very high
14
penetration, and we can get these pretty amazing
15
responses in patients who have significant
16
leptomeningeal disease.
17
symptomatic, and we can see these really nice
18
responses.
19
And what we've seen is that alectinib
They may be very
However, they can relapse as well.
This particular patient as treated at the
20
recommended phase 2 dose of 600 milligrams twice a
21
day, and he relapsed after just 4 months.
22
this situation, he relapsed at very significant
A Matter of Record (301) 890-4188
And in
250
1
leptomeningeal disease again, really no other
2
treatment options.
3
the FDA, we did dose escalate this patient to a
4
higher dose, the 900-milligram dose, and this
5
patient actually did re-respond, resolution of his
6
symptoms, radiological response, and actually has
7
now been on an additional 5 months.
8
wonder whether or not 600 milligrams twice a day is
9
going to be the right dose for all of our patients.
10
Again, just to summarize the alectinib dose
11
finding, this study used a modified 3 plus 3 design
12
and did identify a recommended phase 2 dose, did
13
not clearly identify a maximum tolerated dose.
14
hundred milligrams twice a day was selected based
15
on a number of parameters:
16
particularly efficacy that was seen at that dose
17
level.
So in agreement with Roche and
So you have to
Six
safety, PK, and
I would say that there were other factors
18 19
probably that went in to choosing 600, including
20
timing.
21
competitive with all those ALK inhibitors out
22
there.
This landscape has become incredibly
There was a real, I think, desire to move
A Matter of Record (301) 890-4188
251
1
this drug quickly into phase 2 and phase 3 studies,
2
and also potentially a drug formulation issue in
3
terms of thinking about that intermediate dose of
4
760.
5
What I've seen in practice is that alectinib
6
is extremely well tolerated at 600 milligrams.
I
7
don't very often have to dose reduce.
8
is really going to help alectinib in terms of
9
thinking about combinations and how we can combine
I think this
10
ALK inhibitors with other targeted therapies, and
11
potentially immunotherapies.
12
mentioned, I really think higher doses for
13
alectinib need to be explored because there are
14
situations, particularly CNS disease, where I think
15
higher doses are going to be very beneficial to our
16
patients.
17
However, as I
Just to conclude -- and actually, I think
18
almost every one of these bullet points has already
19
been touched on in the last two days.
20
conclude, we've seen in the ALK landscape that
21
different dose escalation designs have been used to
22
select the proper doses, what we hope are the
A Matter of Record (301) 890-4188
But just to
252
1
proper doses for different ALK inhibitors.
2
We've already talked about how we really
3
need to look at toxicities beyond just the first
4
cycle of treatment.
5
illustrates that well.
6
illustrates that the impact of food on PK and
7
toxicity really needs to be explored prior to
8
settling on the MTD or recommended phase 2 dose.
9
would really have liked to have the opportunity to
I think the ceritinib story Again, ceritinib
10
explore different dose levels, either within the
11
phase 1 expansion or in the phase 2 trials.
12
think that would really have helped us in terms of
13
really fine tuning the correct doses.
14
I
Just the last thing I'll mention, which came
15
up on day 1 yesterday, which that there are these
16
rare but life-threatening AEs that are associated
17
with ALK inhibitors, particularly pulmonary
18
toxicity.
19
really need to do more work on in the preclinical
20
arena to understand who's really at risk for that.
21 22
I
And I feel like that's one area that we
Finally, I just acknowledge the patients. would say, in particular, they go through these
A Matter of Record (301) 890-4188
I
253
1
dose escalation studies, which are really, really
2
time consuming and logistically really challenging
3
for them, and they really put up with a lot to
4
allow us to try and develop these drugs.
5
course, all of the investigators on the ALK
6
inhibitor studies have been incredibly invested and
7
helpful.
8
have developed the drugs.
9 10
Of
And finally, of course the sponsors who Thank you.
(Applause.) DR. SHAW:
I'll just introduce the last
11
speaker.
12
last speaker is Dr. Pasi Janne from Dana Farber
13
Cancer Institute in Boston.
14 15
And again, thank you for staying.
The
Presentation - Pasi Janne DR. JANNE:
Thank you.
I wanted to spend
16
some time today talking a little bit about
17
combination strategies and really focus my
18
discussion on EGFR mutant lung cancer, where we
19
really struggled for a long time trying to develop
20
combination approaches, and hopefully some of these
21
comments that will be applicable to many of the
22
other indications.
Here are my disclosures.
A Matter of Record (301) 890-4188
254
Certainly, there is a lot of preclinical
1 2
rationale to develop combinations of kinase
3
inhibitors.
4
that single agents aren't going to cure our
5
patients with advanced cancer of any kind.
And I think we're all keenly aware
Several different kinds of combinations and
6 7
good preclinical data have suggested that, for
8
example, combining EGFR inhibitors with MET
9
inhibitors, or MEK inhibitors, or combining PI3
10
kinase inhibitors with MEK inhibitors could be
11
potentially therapeutically successful.
12
problem is that many of these work in, of course,
13
preclinical models until you are successful in the
14
clinic.
15
couple of days about many of the challenges in this
16
field.
17
The
And we of course have heard over the last
Some of the challenges are, again, as we've
18
heard over the last couple days, that many single
19
agents are still developed on an MTD model.
20
They're administered continuously.
21
are long half-life drugs, are often prioritized in
22
preclinical development into the clinic, and many,
A Matter of Record (301) 890-4188
They're often
255
1
of course, of our TKIs have overlapping toxicities,
2
and it's not clear that in many instances, there's
3
a clear therapeutic window, especially when you
4
start to do combination therapies.
5
If you look here, for example -- and again,
6
I'll make many of the comments on the EGFR front.
7
So erlotinib would be a good surrogate here,
8
whereby if you try to develop combinations with
9
erlotinib, if erlotinib is looked at as drug
10
number 1, and you combine it with drug number 2
11
that has potentially many of the same
12
toxicities -- rash, diarrhea, elevated LFTs,
13
occasional pneumonitis -- when you develop these in
14
combination, you run into intolerable toxicity.
15
And the question really is, how can one get around
16
that, and are there some strategies to potentially
17
overcome this problem?
18
Let me show you some examples of where we
19
have effective therapies, but they are toxic.
20
Again, in the EGFR space, the combination of
21
afatinib and cetuximab, very nice preclinical data
22
in the EGFR T790M model.
This is the most common
A Matter of Record (301) 890-4188
256
1
reason for developing resistance to erlotinib and
2
geftinib and afatinib, is the development of EGFR
3
T790M.
4
very good strategies to overcome this.
5
preclinical data suggests that the combining of
6
afatinib and cetuximab is a potentially effective
7
strategy presumably because the combination leads
8
to degradation of EGFR.
9
driven by a mutant oncogene, degrading the mutant
And until recently, we didn't really have And this
And in a cancer that's
10
oncogene is essentially as good as inhibiting it
11
fully.
12
This was taken forward into the clinic, and
13
this is a phase 2 clinical trial that's
14
subsequently been published in Cancer Discovery,
15
where you can see the response rate of 30 percent,
16
a PFS of 4.7 months.
17
came out, there really was very little -- or any
18
single agent had less than 10 percent response rate
19
in this patient population.
20
a step forward.
21 22
And really, until this data
So this was definitely
Now, the problem comes in the tolerability. This may be somewhat hard to see.
A Matter of Record (301) 890-4188
These are the
257
1
individual patients and the patients that have the
2
T790M mutation or don't have the T790M mutation.
3
And in dark green are the patients who had a PR,
4
and then in light green are those who had to dose
5
reduce.
6
response was stable disease, and light blue is when
7
they had to dose reduce. You can see the vast majority of individuals
8 9
And in darker blue are patients whose
here could not tolerate the combination for more
10
than a few months at most, and then had to dose
11
reduce.
12
adverse events in this clinical trial, the grade 3
13
treatment related AEs were at 44 percent, and the
14
grade 4 treatment related AEs were at 2 percent.
15
And if you look at the treatment related
Most of these are on-target toxicities.
16
Both afatanib and cetuximab are great inhibitors of
17
wild type EGFR, and really, the combination of
18
these two is a very good way of inhibiting
19
wild type EGFR and leads to really intolerable skin
20
toxicity in most of the patients.
21
really the most common reason for dose reduction
22
here, although diarrhea was also seen as well.
A Matter of Record (301) 890-4188
And that was
258
1
Another combination that both myself and
2
Alice actually were involved in, that was to
3
combine dacomitinib and crizotinib.
4
dacomitinib, the covalent EGFR from Pfizer, which
5
was originally thought to be able to overcome EGFR
6
T790M in patients, subsequently was found out that
7
wasn't the case.
8
crizotinib not because of its use as an ALK
9
inhibitor, but because of its use as a MET
Here
And here, combined with
10
inhibitor, T790M and METs are two common ways in
11
which lung cancers can develop resistance to
12
erlotinib.
13
phase 1 combination trial to see if it would be a
14
good combination strategy to overcome resistance.
15
You can see here the different doses that
16
were taken forward, the DLTs that happened at the
17
different dose levels.
18
some expansion cohorts that were performed, but at
19
the end of the day, the combination was both not
20
particularly efficacious because dacomitinib
21
ultimately could not inhibit T790 in patients, and
22
unfortunately toxic.
And again, the idea was to do this
And ultimately, there were
A Matter of Record (301) 890-4188
259
1
If you look at the specific toxicities of
2
the expansion cohorts and escalation cohorts, and
3
then look at over here the summation of the
4
toxicities, again, you see a significant percentage
5
of patients, 44 percent of patients with grade 3
6
treatment related toxicities, and a few percent of
7
individuals with grade 4 treatment related
8
toxicities. The issue here is that if we're going to
9 10
develop combination therapies for patients, then we
11
expect of hope that these are treatment strategies
12
on which a patient's going to be on a therapy for
13
extended periods of time, not just weeks but maybe
14
months or years, it has to be one that is
15
ultimately tolerable.
16
demonstrate that it's not feasible.
And these two examples
Again, skin toxicity, the most common
17 18
toxicity from EGFR kinase inhibitors, and it is
19
debilitating for patients over a long period of
20
time, over months to years on patients who are on
21
it.
22
to dose reduce or often dose interrupt.
This is the most common reason patients have
A Matter of Record (301) 890-4188
260
1
Other combinations, I mentioned PI3 kinase
2
and MEK inhibitors, again, elegant work looking at
3
this, for example, in KRAS mutant lung cancer in
4
preclinical models.
5
mouse models, the combination's quite effective.
6
It inhibits signaling pathways, and many, many
7
combinations are ongoing or have been completed.
8
And again, the same issues run into play here, that
9
given these therapies, which inhibit signaling
In genetically engineered
10
pathways, which are not just critical for cancers
11
but are important for other normal homeostatic
12
processes in life, you probably don't have the
13
therapeutic window here to be able to fully inhibit
14
these pathways of cancers without having side
15
effects in patients.
16
How can one overcome some of these issues or
17
what kinds of strategies can we at least think
18
about this?
19
dosing schedules.
20
continuously, giving them in an intermittent way or
21
some other methods.
22
One is to think about alternative As opposed to giving things
We think of erlotinib, for example,
A Matter of Record (301) 890-4188
261
1
erlotinib is developed on an MTD model,
2
150 milligrams once a day in a continuous fashion
3
is the registered dose, which is what most patients
4
start on for their EGFR mutant lung cancer.
5
However, there is data on giving erlotinib at
6
25 milligrams a day, in fact, a case series that
7
was published several years ago suggesting that if
8
you have EGFR mutant lung cancer, this is an
9
effective dose to give to individuals, and
10 11
certainly much better tolerated. If you recall, erlotinib was not developed
12
as an inhibitor of mutant EGFR.
13
as an inhibitor of wild type EGFR.
14
probably a hundred-fold window or a therapeutic
15
window over the dose required to inhibit mutant
16
EGFR at the 150-milligram dose.
17
25-milligram dose, you can effectively inhibit
18
mutant EGFR and not see the same degree of
19
toxicity.
20
that have taken this forward as a combination
21
partner.
22
It was developed And there's
And hence, at the
But there are very few, if any, studies
You can also give erlotinib once a week, and
A Matter of Record (301) 890-4188
262
1
this is commonly used for patients who have the
2
same problem that Alice presented, which is
3
leptomeningeal disease, whereby penetration into
4
the CSF is a major issue for patients who develop
5
that progression or develop isolated CNS
6
progression and who have systemic control of their
7
disease.
8
week.
Oftentimes they're given erlotinib once a
Sometimes this is in fact much better
9 10
tolerated because you have a transient high dose of
11
erlotinib, and then have several days of the week
12
for normal tissues to recover from that inhibition
13
because erlotinib doesn't have a particularly long
14
half-life; again, strategies that one could
15
incorporate using this approach in a combination
16
type strategy. There is one clinical trial ongoing that's
17 18
trying to do a little bit of a mixture of both of
19
these.
20
by Dr. Helena Yu at Memorial Sloan Kettering.
21
this is a trial based on preclinical data,
22
suggesting that pulsatile erlotinib followed by
This is a clinical trial that's being run
A Matter of Record (301) 890-4188
And
263
1
low-dose continuous erlotinib could be a strategy
2
that would be, number one, better tolerated, and in
3
the preclinical space could be at least suggested
4
that it would inhibit the emergence, again, of the
5
EGFR T790M mutation.
6
You can see here the different dose levels,
7
where the erlotinib is given two days in a row at a
8
higher dose, and then at a lower dose the rest of
9
the week.
This is an ongoing study.
I don't have
10
any updates in terms of the data on it, but it's
11
definitely enrolling patients at this point.
12
There are also different properties, again,
13
of inhibitors.
So for example, if you look at MEK
14
inhibitors, trametinib compared to selumetinib, you
15
have different half-lives of these agents.
16
certain combinations, having a short half-life,
17
partner drug allows intermittent dosing, so you
18
don't have to do continuous dosing versus where
19
that may be a challenge for a drug that has a 5 and
20
a half day half-life.
21
do intermittent dosing or even combination dosing
22
because dose reductions, by the time you clear the
For
It may be more difficult to
A Matter of Record (301) 890-4188
264
1
drug or you get to a level where you know it's no
2
longer detectable, it can be several days.
3
We saw this earlier, again, from a clinical
4
trial that's ongoing that Dr. Shapiro presented at
5
ASCO; again, different types of strategies of when
6
you do combinations of dosing one agent and one
7
cohort elevated, or starting in a cohort where
8
you're dosing intermittently here, et cetera, and
9
ultimately to achieve potentially a dose that's
10
both tolerable or a combination of doses that's
11
both tolerable and efficacious.
12
I'll switch to the -- the next topic is what
13
can we do in terms of our drugs?
Our drugs are
14
something that we can modify to dial out the
15
toxicities.
16
different types of VEGF inhibitors, for example,
17
panatinib, which has a wide spectrum of kinases
18
that it inhibits.
19
different ALK inhibitors, for example, crizotinib,
20
that inhibits many, many more targets than just
21
ALK, compared to alectinib, which is much more
22
selective.
We heard earlier on yesterday about
You heard from Alice about
A Matter of Record (301) 890-4188
265
1
In the EGFR space, toxicity from EGFR
2
inhibitors comes from on-target toxicities, comes
3
from inhibiting wild type EGFR.
4
EGFR is not something -- or inhibition of wild type
5
EGFR is not something you need for an EGFR
6
inhibitor to be efficacious in EGFR mutant cancers.
7
They're solely driven or dependent on mutant EGFR
8
and not on wild type EGFR.
9
And yet, wild type
The hope could be that you could
10
differentiate these two or selectively inhibit one
11
over the other, and that way potentially dial out
12
the toxicity that we so commonly face with EGFR
13
inhibitors, which really prevents a lot of the
14
combination studies from being done.
15
Several years ago, we tackled this issue and
16
published this preclinical paper to suggest that
17
this could be possible by identifying a novel class
18
of EGFR inhibitors, which now differentially
19
inhibited mutant versus wild type EGFR.
20
see here these agents are in this IC50 graph,
21
incredibly potent against the cell line model, but
22
has a deletion T790M.
You can
This is in vivo inhibition
A Matter of Record (301) 890-4188
266
1
of that.
2
At the same time, this tool compound was
3
completely ineffective in a model, the A431 cell
4
line model, which is a model that is a wild type
5
EGFR amplified and where the growth of that model
6
is completely dependent on the EGFR wild type
7
kinase activity.
8
which inhibits, again, both mutant and wild type
9
EGFR activity.
It is sensitive to erlotinib,
And here's just to show at the
10
level, that you don't inhibit phosphorylation of
11
EGFR until you have about a hundredfold window in
12
terms of not being able to inhibit that. Again, to try show this graphically, if you
13 14
look at -- and this is just a representative
15
example of this.
16
gefitinib or erlotinib, they're very good
17
inhibitors of mutant EGFR.
18
of wild type EGFR.
19
does with the kinase domain, you cannot inhibit
20
T790M.
21
inhibitor of mutant EGFR, it's a better inhibitor
22
of the T790M, but it's also a better inhibitor of
If you look at a current EGFR, so
They're good inhibitors
And because of what the T790M
If you take afatinib, which is a good
A Matter of Record (301) 890-4188
267
1
wild type EGFR.
And again, this is one of the
2
reasons:
3
inhibit T790M in patients because you're limited by
4
on-target toxicity.
with afatinib, you are not able to
The next generation EGFR inhibitors have
5 6
this split.
They're much better inhibitors of the
7
mutant in wild type, and that includes inhibition
8
of the resistance mutation.
9
this property of mutant selectivity of being able
10
to dissociate the ability to inhibit wild type in
11
mutant EGFR.
And again, this gives
Now, several of these compounds are in the
12 13
clinic.
14
in the clinic.
15
along is the AstraZeneca AZD9291 and the
16
Clovis CO-1686 or rociletinib.
17
been granted FDA breakthrough designation, and you
18
can see the others are here.
19
out at ASCO on several of these in just a few
20
weeks.
21 22
There are at least seven of them that are The ones that are the furthest
Both of them have
There's data coming
We've been involved extensively with the AstraZeneca compound, and again, I'll take you
A Matter of Record (301) 890-4188
268
1
through the design of the trial and, again, how the
2
doses were chosen moving forward.
3
phase 1 trial of the AZD9291 compound that was
4
specifically for patients that had failed prior
5
EGFR inhibitors.
6
Rolling 6 design, so not a 3 plus 3 design, so
7
6 patients were entered in each cohort.
The initial cohorts -- this was a
If no DLTs were observed in any of the
8 9
This was a
cohorts, two things happened simultaneously.
A
10
subsequently higher cohort could be opened, and at
11
the same time an expansion cohort was opened.
12
idea of an expansion cohort was to then explore the
13
possibility is there activity.
14
not have to have toxicity, you had to have some
15
hint of activity before the expansion cohort
16
opened.
17
to then evaluate did the agent have activity
18
broadly in patients that developed resistance to
19
EGFR inhibitors and was that activity
20
differentiated in patients that had the T790 and
21
the common resistance mechanism versus those that
22
did not.
The
So not only did you
And the idea in the expansion cohorts was
A Matter of Record (301) 890-4188
269
1
Here, patients needed to be at central
2
confirmation and central testing of their EGFR
3
mutation status.
4
were treated, or patients were treated on 5 dosing
5
cohorts.
6
was identified in this trial, and patients in all
7
of these cohorts, again, were treated in these
8
expansion cohorts and, again, evaluated for
9
efficacy.
And ultimately, 5 dosing cohorts
There were no DLTs identified and no NTD
10
If you look across the -- this is now just
11
the T790M population of patients, you can see that
12
there is activity.
13
seen in all 5 dosing cohorts.
14
rates of responses are seen in all 5 dosing
15
cohorts.
16
efficacious doses here.
17
is that in the patients that don't have the T790M
18
mutation, the response rate is much lower and the
19
PFS is significantly shorter in the patients that
20
don't have the T790M mutation compared to those
21
that do not.
22
And the activity is actually And roughly, similar
So you in fact have 5 potentially What I don't show you here
What about toxicity?
Here -- and maybe
A Matter of Record (301) 890-4188
270
1
you'll be able to see.
2
into patients that have AEs or AEs that are felt to
3
be drug related, and here are grade 3 or higher AEs
4
that are felt to be drug related.
5
sense that there is some increase based on dose and
6
more, I think, better.
7
we've now separated the different types of
8
toxicities.
9
So the toxicity is divided
And you get some
You can see here where
Now, if you look at these toxicities here,
10
which are most reflective of EGFR-based toxicities,
11
again, on-target, EGFR-based toxicities, certainly
12
once you start to get to the higher doses, the 160-
13
and the 240-milligram dose, you start to get a
14
sense of a greater degree of these toxicities, not
15
very many grade 3's but certainly grade 1 and 2
16
toxicities.
17
interested in developing agents here that are
18
tolerable for a long period of time, and even
19
grade 2 skin toxicity or grade 2 diarrhea is not
20
something that is desirable.
21
of course, they're not exclusive in their ability
22
not to inhibit wild type EGFR; they're just
And again, remember, we were
This class of agents,
A Matter of Record (301) 890-4188
271
1
selective.
2
start to see this.
3
And again, at the higher doses, you
Based on the toxicity information and the
4
efficacy information, the 80-milligram dose was
5
chosen here to move forward into the next phases of
6
the clinical trials, which include a dedicated
7
phase 2 clinical trial, which has been completed,
8
as well as an ongoing randomized phase 3 clinical
9
trial compared to chemotherapy, and an ongoing
10
randomized phase 2 trial compared to
11
first-generation EGFR TKIs and the EGFR TKI naive
12
patient population.
13
I will say that the preclinical studies
14
predicted that 20 milligrams could be an
15
efficacious dose, which it was.
16
you're getting essentially efficacious doses in all
17
the models that were tested.
18
I think goes into that decision is, again, similar
19
to what Alice mentioned, is CNS penetration.
20
At 80 milligrams,
And the last bit that
One of the issues we have with the patients
21
with EGFR mutant lung cancer as well as CNS
22
relapse, there are certainly isolated cases of
A Matter of Record (301) 890-4188
272
1
individuals who have had brain metastases on entry
2
where they've been treated that have regressed on
3
9291.
4
things are planned.
5
a higher dose with the ability to penetrate the
6
CNS.
7
It hasn't been a dedicated study; those But the idea also was to have
There are now, as I mentioned, several of
8
these agents out there, and now there are
9
combination trials ongoing with many of them.
And
10
this probably doesn't capture all of them.
So
11
again, with the AZ drug, immunotherapy combination,
12
now a MET combination as well as a MEK combination;
13
with the rociletinib, a MEK combination,
14
immunotherapy combination; Aurora kinase
15
combination, immunotherapy combination, MET
16
combination.
17
So similar themes emerging here.
18
hope here, of course, is that we're now able to
19
develop tolerable combinations for patients because
20
we don't have potentially all those overlapping
21
toxicities that we have with afatinib and cetuximab
22
or with dacomitinib.
A Matter of Record (301) 890-4188
And the
273
The real challenge, next comes is which
1 2
combination therapy should be used and when should
3
it be used in patients who develop resistance of
4
this class of agents, should it be used as a
5
combination strategy to overcome erlotinib
6
resistance, or should these ultimately be used in
7
patients who are TKI naive but who have EGFR mutant
8
lung cancer if indeed they are combinations that
9
are tolerable?
And I think this is next phase of
10
both testing in preclinical models to predict that,
11
as well as in the next phases of clinical trials. Some future considerations, a mutant
12 13
selective inhibitor approach, which was successful
14
here.
15
other oncogene-driven cancers where you could
16
potentially take advantage of a mutation that's
17
cancer specific.
18
One could potentially implement this in
Combination strategies, again we talked
19
about intermittent dosing.
I think one of the
20
other things that we don't do very well, at least
21
in the lung cancer community, is do PD in-human
22
tumors, actually biopsy patients' tumors and ask
A Matter of Record (301) 890-4188
274
1
how much and for how long do you need to inhibit
2
the target for things to be effective.
Are you
3
actually inhibiting your target fully?
Is it that
4
you're inhibiting your target and the cancer
5
doesn't shrink, or are you not inhibiting the
6
target, and the cancer doesn't shrink?
7
those are two different answers and lead you down
8
two different paths, and something that I think we
9
need to continue to do more work.
10
I think
Other considerations, the hope is, of
11
course, the combinations will lead to long-term
12
efficacy.
13
not be appropriate, and maybe one needs to think
14
about other things for combination therapies, what
15
fraction of patients completed X months of therapy
16
with the dose or the doses of the agents that they
17
started in?
18
dose reductions or interruptions?
19
We talked about the DLT windows that may
What fraction of patients required
The afatanib/cetuximab example is a good one
20
because I think you don't see too many individuals
21
that were able to complete that because of
22
toxicity.
And I think, again, the hope is that
A Matter of Record (301) 890-4188
275
1
we'll have therapies that last for a long period of
2
time, so again, they have to be tolerable.
3
that, I'll stop.
Thanks.
(Applause.)
4 5
So with
Panel Discussion - Geoffrey Kim and Alice Shaw DR. KIM:
6
At this time, we'd like to invite
7
all the panelists for the final presentation the
8
podium.
9
want to do in this session is to really get the
While you gather, I think, really, what we
10
take-home message.
11
and every one of the panel members, what is the
12
take-home message.
13
start thinking about what your response would be.
14
So we're going to ask you, each
So as you gather to the podium,
I believe we have only one new member on the
15
panel -- or two new members on the panel that need
16
to be introduced.
17
Dr. Barrett, can you introduce yourself?
18
DR. BARRETT:
Jeff Barrett, Sanofi
19
Pharmaceuticals.
20
pharmacometrics program.
21
DR. KIM:
22
I'm the head of interdisciplinary
Dr. Yates, also can you introduce
yourself?
A Matter of Record (301) 890-4188
276
DR. YATES:
1 2
I'm a modeling and simulation
scientist working preclinically, AstraZeneca.
3
DR. KIM:
Great.
4
I'll go ahead and start about what's the
5
take-home message.
And I think we've learned a
6
tremendous amount, and I really want to express my
7
gratitude to all the sponsors who were very
8
transparent with a lot of the processes that they
9
have.
I think you really can see the passion for
10
what you do, and it's evident in the quality of the
11
work that has been presented during this workshop.
12
I think we all are faced with the challenge of
13
doing better, and I think we all agree that we can
14
do better.
15
One of the aspects that struck me was
16
actually this session with Dr. Petruzzelli talking
17
about education, about taking that model concept
18
and really sitting down internally first, and then
19
also externally, and educating about what the
20
benefit of implementing this type of strategy would
21
be, both internally and externally.
22
that's really one important concept that I want to
A Matter of Record (301) 890-4188
And I think
277
1
take home from this, that internally at the FDA, we
2
certainly could benefit from more education
3
internally about the utility of these models.
4
But at the same time, I do think that we are
5
all waiting for a success story using an adaptive
6
design in dose finding, so that we can point to
7
that story as kind of a key lesson to be learn and
8
why this path should be emulated for drug
9
development.
10
So we look forward to -- especially as we
11
enter the world of combination therapies -- and it
12
was a perfect segue into this discussion because I
13
think we're going to see a lot more of those
14
combination studies.
15
integrate the knowledge that we have about
16
preexisting compounds or compounds in the pathway,
17
that we're really going to be stuck with an
18
inefficient trial design.
19
And if we really don't
So I think the take-home message is really
20
more education internally, externally, and being
21
able to communicate the benefits of using novel,
22
innovative, adaptive designs.
A Matter of Record (301) 890-4188
So let's go down the
278
1
line and, Dr. Shaw, perhaps you could speak briefly
2
about the lessons you learned, if any, here at this
3
workshop.
4
DR. SHAW:
This is great.
I think what came
5
out loud and clear, and we've talked about this
6
yesterday and today, is that collaboration among
7
multiple groups is really essential for, I think,
8
optimizing the doses, optimizing the design, and
9
then ultimately identifying the right dose.
And
10
we're talking about collaborations between not just
11
the sponsors and of course the clinical
12
investigators, but also the FDA as well.
13
I think this will become particularly
14
important when we think about the combinations.
15
Dr. Janne was talking about these combination
16
studies are going to be so complicated, and I think
17
there, collaboration is going to be essential to
18
get these combinations off the ground.
19
DR. KIM:
Dr. Janne?
20
DR. JANNE: I would echo that.
I think that
21
message came through very clear from looking at
22
both clinical data, preclinical data, PK data to
A Matter of Record (301) 890-4188
279
1
try to understand what may be optimal dosing.
And
2
we don't always have the right dose when we enter
3
the clinic, and you learn, obviously subsequently
4
from the clinic there, what may be -- with
5
toxicities that develop, et cetera. I think one of the other things that was
6 7
clear or something that was new to me was to really
8
start thinking about evaluating multiple
9
efficacious doses, even in advance when you're in
10
the phase 2 and phase 3 setting because you don't
11
always know from your phase 1 what the optimal dose
12
may be.
13
big one and feeding back information on both the
14
preclinical and clinical sides.
15
DR. PETRUZZELLI:
But certainly communication I think is the
I would -- again, there's
16
going to be a lot of echoing.
But I think the main
17
take-home for me is how we're going to have to
18
learn from our experiences.
19
of an iterative process, and we are in a learning
20
phase, particularly if we switch from what we were
21
using before, to any new models.
22
there are lots of factors that we learned, that
I think this is sort
A Matter of Record (301) 890-4188
But in the end,
280
1
some of the models are going to be important and
2
helpful, and some of them are just the data coming
3
in, and how do we constantly communicate and learn
4
from each other as we move forward to become more
5
efficient. DR. BAILEY:
6
I think, Geoff, you asked for
7
case studies.
I think the last two presentations
8
presented four case studies.
9
take crizotinib, the 3 plus 3.
To that point, let's Although it's a
10
3 plus 3, other information was used to end up
11
making a decision on the dose.
12
already talked about.
13
The alectinib clearly wasn't just a simple 3 plus
14
3.
15
design to get additional information, the dose
16
levels, and the AZD, the Rolling 6, again, it
17
wasn't just DLT data.
18
Ceritinib we
I won't echo that one again.
We had the flexibility built into the study
It doesn't matter whether you're using a
19
model or an algorithm.
It's about flexibility that
20
you build in to be able to collect the other
21
information.
22
model is really what happens if something goes
And your final decision to use a
A Matter of Record (301) 890-4188
281
1
wrong.
2
modeling helps, but building flexibility is the
3
key.
4
And I think that's where potentially the
DR. BARRETT:
I guess one of the things that
5
struck me as I see a lot of the presentations here
6
is that we talked about targeted therapy, and in
7
many of the slides, we saw a single line
8
representing some target exposure.
9
change for patients, it didn't change over time.
10 11
It didn't
That's fundamentally flawed. I think the other problem that we have is
12
that in this patient population, we're assuming
13
that these people have the same probability of DLT
14
for every dose cohort, and that that's translatable
15
to the next study.
16
difficult assumptions to make.
17
I think both of those are very
I think fundamentally, we have to think a
18
little bit differently about the target existing
19
within the patient.
20
session off by talking about the high rate of dose
21
reductions and discontinuations in clinical trials
22
projecting into reductions in real-world adherence.
Geoff, you started this
A Matter of Record (301) 890-4188
282
1
I think fundamentally none of what we discussed is
2
really getting at that.
3
associated with the way we're studying this and the
4
assumption that we can come up with this optimal
5
dose, optimal dose for who?
6
I think probably this is
What you saw from Dr. Shaw's presentation
7
was very compelling in that she has individual
8
patients that are making their own decisions.
9
in this era of social media, this information will
And
10
become more and more public, and we're going to
11
have to educate not just our own community of
12
practice here, but beyond that because the patients
13
are going to ask for more.
14
So I think fundamentally we have to think a
15
lot more about how we're going to characterize this
16
target population, be able to make some a priori
17
assessment of the vulnerability of the likelihood
18
of an individual patient developing toxicity
19
because they're not independent.
20
got to be part of the future epiphany.
21 22
DR. JIN:
I think that's
I think what to me is most
impressing is how common we all are.
A Matter of Record (301) 890-4188
We all share
283
1
the same challenges.
2
encouraging for me, from all the regulators,
3
industry partners, and also the clinical folks,
4
that we all share the similar challenges in
5
oncology about toxicity, long-term safety, how to
6
basically identify effective dose very efficiently
7
and quickly, how to do the dose optimization.
8 9
It's actually probably very
It's also very eye-opening to see different people have taken on different initiatives or
10
explore different ways other than the traditional
11
approach, adaptive designs for identifying
12
different patient population, the dose optimization
13
based on different endpoints for doing the
14
dose-ranging studies, for having food effect or
15
dose titrations, different ways of getting to a
16
better benefit quickly and also manage
17
tolerability.
18
I think that aspect is very eye-opening.
19
It's really more about the cross-learnings.
20
Hopefully, we can take all this information back
21
and think how we can integrate all this information
22
and do things more smartly, everyone.
A Matter of Record (301) 890-4188
284
DR. McKEE:
1
From the agency perspective, I
2
think what's been very illuminating for many of us
3
is how much thought goes into this that we never
4
see because it doesn't make it into regulatory
5
submission.
6
justification for the selected recommended phase 2
7
dose.
Basically, we see a two-paragraph
So I think this should become more of a
8 9
conversation between the agency and the sponsors
10
because, clearly, we're very interested in this,
11
and we want to know more about why you came to this
12
dose.
13
clinical investigators.
14
a lot about the PK/PD data.
We don't hear about the input from the We don't necessarily hear
15
So I think we would like to hear more about
16
that, and not from a dose justification, but from a
17
dose explanation.
18
that you have, that you could tell us why you
19
selected this dose.
We jus want to know what data
20
I think the other thing that really struck
21
me was at the end, the clinical presentations from
22
our investigators, that maybe these doses that are
A Matter of Record (301) 890-4188
285
1
being selected still aren't the right ones when you
2
get into the real-world population when you've used
3
a DLT window that is relatively brief. I know there are limitations to phase 1
4 5
studies with the patients that are enrolled in
6
those trials, but I think we really need to
7
consider when we move these drugs into the real
8
world, that we may not have the optimal dose still,
9
and how we need to figure this out, whether it's
10
using slightly less perfect, less healthy patients
11
in our trials or try to incorporate more data into
12
how we would treat these patients from the
13
postmarketing perspective.
14
DR. RATAIN:
Well, I'm struck by some just
15
inherent inconsistencies in what we're doing here.
16
We're hearing about drugs that are particularly the
17
latest generation that are highly selective
18
compounds, yet we're still pushing them towards the
19
MTD.
20
going to be highly variable pharmacokinetics, yet
21
we're not seeing pharmacokinetic data in the
22
context of trying to understand relationships
These are oral drugs.
We know that there's
A Matter of Record (301) 890-4188
286
1
between efficacy and PK variability or safety and
2
PK variability.
3
So I think that we really need to think more
4
about this.
I mean, if these drugs do have truly
5
narrow therapeutic indices, then we have to expect
6
that at the MTD, there's going to be so enough PK
7
variability that's going to be a major driver of
8
safety and efficacy, or they're not narrow
9
therapeutic index, in which case we should be
10
backing off from the MTD.
11
DR. YATES:
I think the standard thing for
12
me has been the idea that a phase 1 trial shouldn't
13
be just about finding MTD but robustly identifying
14
the exposure-response relationship or safety and
15
hopefully efficacy as well.
16
important is it puts you in a good place to start
17
thinking about combinations because a single-agent
18
curve is a prior for the combination.
19
The reason that's
I think the use of preclinical data and the
20
different ways that different companies are using
21
it is quite interesting and refreshing as well.
22
And I agree with the comment that a single line and
A Matter of Record (301) 890-4188
287
1
a graph probably isn't realistic.
2
certainly going to take back is how can we build
3
variability and how can we design preclinical
4
experiments to assess and reproduce interpatient
5
variability, maybe not concentrating on a large
6
number of animals in single models, but looking at
7
PD access and running -- as it relates to before
8
the idea of clinical trials in mice, implanting
9
different models in different mice and looking for
10
a covariate in terms of wild type versus mutations
11
on the pathway of interest. DR. McKEE:
12
A challenge I'm
And from a nonclinical
13
perspective, I think my major take-aways here are
14
something that came up yesterday and today.
15
thing is how very important it is for our clinical
16
colleagues, as they gather these data on PK/PD, to
17
come back and have focus discussions with their
18
preclinical colleagues to really understand how to
19
best use the toxicity models and the efficacy
20
models.
21 22
One
I think this is going to become even more important when you want to try to establish
A Matter of Record (301) 890-4188
288
1
combination therapies and long-term dosing because
2
on each individual molecule basis, if we have that
3
discussion, then I think we can have a really
4
rational approach to the best way to use the
5
preclinical models, or the efficacy, or especially
6
toxicity models, in a focused way so that it's not
7
a blanket statement of let's just combine these
8
drugs.
9
what you want to see from those models.
But you come back with an hypothesis about
I know we certainly have those discussions,
10 11
and they haven't always panned out in terms of us
12
running the models, i.e., we haven't helped the
13
clinic.
14
a good way to have those discussions and try to
15
integrate that preclinical data.
16
But I'm reaffirming that this is probably
DR. KIM:
I definitely think one of the
17
common themes is the communication issue.
What do
18
you think the major barriers to communication are?
19
Is it just that it hasn't been thought of before or
20
are there real barriers that need to be overcome in
21
the access?
22
clinical and especially the investigators with the
Is there perhaps no access between the
A Matter of Record (301) 890-4188
289
1
pharmacometric and the pharmacocologist to
2
communicate this type of discussion that Dr. Ratain
3
brings up?
4
that the flow of information is not moving
5
optimally?
6
limitations?
7
who's willing to answer; otherwise, I'll pick on
8
people.
Is there a disconnect of information
And then how can we overcome those I'll open up the floor to anybody
9
Dr. Ratain, in your experience as an
10
investigator, do you ever get the feedback that you
11
need or desire regarding the exposure where the PK
12
and the modeling falls out, and where do you think
13
the therapeutic index occurs?
14
necessary information that you desire? DR. RATAIN:
15
Are you getting the
In the trials we do with
16
industry, I would say that the data does not come
17
back -- the PK's data not come back on any regular
18
basis.
19
it's not as consistent as I would like.
20
know whether that's because there's a lag in which
21
the data are analyzed or if it's really a
22
communication barrier.
Some trials are better than others.
A Matter of Record (301) 890-4188
But
I don't
290
DR. KIM:
1
I believe, Dr. Shaw, you brought
2
up that you see a signal, a potential signal, with
3
pulmonary toxicity.
4
communicate with perhaps the industry, nonclinical
5
people to say, hey, this is a potential signal here
6
that I'm seeing; do you have a model for this, or
7
have you investigated something like this? DR. SHAW:
8 9
up.
Were you able ever to
Yes.
That definitely has come
And I agree, there always does seem to be
10
somewhat of a lag in terms of being able to obtain
11
that PK data, so it will come up on telecons.
12
would say the clin pharm people are not often on
13
the telecons, and so it's discussed we need this
14
data; we'll come back to it.
15
we don't come back to it, so unfortunately, it can
16
fall through the cracks.
17
I
And then oftentimes
But I feel like more and more now we're
18
getting better at this and recognizing that that
19
data is so important to us.
20
we're having more success these days on the ongoing
21
trial trying to gather that information to help
22
explain things like unexpected or severe
So I'm feeling like
A Matter of Record (301) 890-4188
291
1
toxicities. DR. KIM:
2
Yes.
I think as clinicians, as
3
the development moves faster and faster, where it
4
behooves us to be more versed in other disciplines
5
and to understand that.
6
curious about the information, and then we start
7
requesting more and more information.
8
opens up the floor as well. DR. JANNE:
9
So I think we become
I think that
I would just add to what Alice
10
said.
I think pulmonary toxicity is one where we
11
don't really have good models, preclinical models.
12
And it's also sort of a black box toxicity,
13
especially -- we both treat lung cancer patients.
14
Sometimes it's hard to separate toxicity from
15
infectious process versus progressive disease.
16
a lot of decisions go into that, but it's one where
17
I think if we had better models or predictive
18
models, it could be very helpful.
So
Audience Q&A
19 20
DR. KIM:
From the audience?
21
DR. BLUMENTHAL:
22
DR. KIM:
I had a question --
Do you want to introduce yourself?
A Matter of Record (301) 890-4188
292
1
DR. BLUMENTHAL:
-- this is Gideon
2
Blumenthal from the FDA -- about the recommended
3
starting dose.
4
ALK inhibitors, for certain patients maybe with CNS
5
meds, you'd want to go higher.
6
Alice, you mentioned with certain
Pasi, you mentioned with erlotinib, it was
7
developed in the wild type at the MTD, and for
8
certain patients, it may be -- for immune patients,
9
you may be able to get by with a lower dose.
And
10
then we heard the example of vandetanib earlier,
11
sort of indolent tumor, medullary thyroid cancer,
12
where maybe a patient might want to go with the
13
lower dose, have less toxicity, maybe have a little
14
bit less response rate, but we don't know how that
15
translates to a longer term time-to-event endpoint.
16 17
I guess when we approve a drug, we usually
18
go with one dose, and that dose is usually the dose
19
in the registration trial.
20
the panel's thoughts on looking at potentially as a
21
recommended starting dose, looking at a range of
22
doses and having maybe the clinician or the patient
I just wanted to get
A Matter of Record (301) 890-4188
293
1
weigh in on what would be the optimal dose for that
2
particular patient.
3
DR. SHAW:
I think that gets to our
4
discussion about how it would be so nice to have
5
studies where we have multiple dose levels tested.
6
I remember for ceritinib, we talked a lot about how
7
to try and incorporate some of the information we
8
had from dose escalation on the lower doses, either
9
into the label or into a publication, to indicate
10
to docs and patients that there is efficacy at
11
lower doses.
12
So for patients who you may feel concerned
13
about starting at the maximum tolerated dose, they
14
would at least feel like there is some guidance or
15
approval to consider a lower dose.
16
really important, and I haven't figured out quite
17
yet how to do it, other than imagining doing these
18
larger expansions of multiple dose levels.
19
DR. BARRETT:
I think that's
I think the challenge is in
20
the logistics to do that because I agree
21
fundamentally with what you're saying, and that
22
would be a better choice.
If we had some
A Matter of Record (301) 890-4188
294
1
intermediate, you could do an effect control trial.
2
It was interesting to me in the morning session, we
3
were almost apologizing for doing dose titration.
4
Why?
5
bad thing?
It's tailored to the patient.
Why is that a
You see the number of strengths that are
6 7
available with the intention of obviously
8
modulating dose as needed.
9
easy to see when you have the PK rationale for
And the PK story is
10
getting within that threshold.
11
doesn't correlate, but you still need to modify
12
dose.
13
But sometimes that
The other situation that, again, we
14
discussed very briefly was over time, you can't
15
expect the dosing requirements to stay the same
16
within patients.
17
what's needed from that standpoint.
18
really about approving a drug that's flexible and
19
where you have options in terms of dealing with the
20
complexities of a disease and the changing
21
complexities within patient.
22
DR. BELLO:
So this too, flexibility is And it's
Akintunde Bello, Bristol-Myers
A Matter of Record (301) 890-4188
295
1
Squibb.
So this is the second time my question has
2
been gazumped, as it was a preceding question by
3
the other questioner.
4
around, say, for the ALK inhibitors, where you
5
essentially almost got two indications -- so you've
6
got the systemic or the organ related lung cancer,
7
and then you've got the brain mets.
8
essentially, again, revisiting this question
9
around, okay, do you need to have two doses or
But around this question
So
10
revisit your strategy for dose evaluation and dose
11
development.
12
The other question or statement I have is
13
around, we're talking about -- a lot of the focus
14
of the discussion is around targeted therapies, and
15
the question around therapeutic index should be
16
wider based on a targeted therapy because you
17
should be relatively selective for your target at
18
the intended site of action.
19
discussed, I think we're seeing variability in the
20
specificity of the target on the cancer versus on
21
endogenous, natural, healthy tissue.
22
But as we've
So there's a question, really, around, okay,
A Matter of Record (301) 890-4188
296
1
when we're talking about targeted therapies, we're
2
talking about various degrees of targeted
3
therapies.
4
the cancer unexpressed anywhere else, then clearly
5
you'll likely to be able to address it with a lower
6
dose.
7
in the cancer, but expressed on some natural
8
tissues that tend to be the sites of toxicity.
First, ones that are very specific to
Then you've got ones that are up-regulated
9
I was just wondering if any of the panel
10
thought about this and thought this was relevant
11
with regards to how we think about dose selection
12
and the question of therapeutic index and dose
13
selection.
14
DR. RATAIN:
Let me respond.
You're right.
15
There are some drugs that you want to push to the
16
max, whether it's oncology or anything else.
17
you want to have some way -- if have a population
18
dose and you don't know what the right dose for the
19
individual is, you give the population dose.
20
you know it could be too high, and you know it
21
could be to low.
22
adverse reactions and you lower the dose.
Then
And
If it's too high, you will see
A Matter of Record (301) 890-4188
297
1
The problem is we don't often know what to
2
do as to whether it could be too low.
And one
3
thought is, well, you could titrate it upward in
4
the absence of toxicity, or we could be more
5
scientific and we can measure plasma concentrations
6
or measure some circulating biomarker.
7
if we're going to use an MTD paradigm and going to
8
start people at the population dose, we have to
9
understand that not only could that be too high for
But I think
10
some patients, it could also be too low for some
11
patients.
12
DR. YATES:
If I could comment about the
13
dose question, might the dose change for the
14
setting, it might be the regimen.
15
to go to intermittent dosing.
16
can butt the optimal PK profile, probably half-life
17
might differ if you want to combine it.
18
want a short half-life compound to combine so you
19
can avoid antagonism, sequence agents, et cetera,
20
allow washout quickly to avoid toxicity. So there
21
might not be an inhibitor for kinase X.
22
on the context it's been used.
A Matter of Record (301) 890-4188
You might want
It might be if you
You might
It depends
298
DR. BARRETT:
1
We have plenty examples of
2
this.
3
methotrexate, where depending on the indication,
4
you have a huge range of doses, and the dosing
5
requirements are known to change within
6
individuals.
7
high-dose methotrexate in various pediatric
8
cancers.
9
If you take a look at the drug like
You're also managing toxicity with
I think there's precedence for this, trying
10
to establish an individual therapeutic window.
11
Again, but you need flexibility and the rationale
12
to be able to make those adjustments and manage the
13
pharmacotherapy not just of the target agent, but
14
in response to other agents that are on board or
15
even rescue therapy.
16
DR. JIN:
I think one thing also, for the
17
target agent platform, if some of the adverse
18
events are target related, that's where I think the
19
cross molecule RNA will be very important.
20
have multiple molecules either in your pipeline or
21
in the public domain hitting the similar target or
22
share the same target subtypes, it's integrating
A Matter of Record (301) 890-4188
If you
299
1
that information to identify what's the
2
tolerability issue, trying to identify what are
3
those tolerability issues that are really target
4
related, and what are some of those that may be off
5
target or the molecule-specific things.
6
those may be helpful information to help maximize
7
the therapeutic window for target agents.
8 9
DR. BAILEY:
I think
I think in terms of the design
of the early dose-finding studies, clearly from the
10
preclinical side where we get estimates of
11
therapeutic index, understanding where we expect to
12
run into safety issues, and understanding, whether
13
it be on target or wild type activity safety
14
issues, and some understanding of where we start to
15
think we're going to see the activity.
16
But yesterday, we clearly heard that the
17
animal models predicting the activity are not as
18
translatable as the models for the safety.
19
we can actually design a trial up front where we
20
say we know that the drug's got a wide or a narrow
21
therapeutic index.
22
drive the escalation through the trial.
So yes,
And that may change the way we
A Matter of Record (301) 890-4188
But if we
300
1
don't have very good efficacy models or
2
translatable models from the efficacy standpoint,
3
then we could be completely misdirected on that.
4
So it gives you a starting point, but reacting to
5
the data you see on the trials is clearly going to
6
be the driver there.
7
DR. GARIMELLA:
My name is Narayana, and my
8
question is about whether in PPA-2 [ph] approaches
9
or combination studies, are they always
10
complementary?
11
incomes?
12
order to have positive outcomes all the time?
13 14
Do they always give us the positive
If not, what can we take as a measure in
DR. KIM:
I had a hard time understanding
your question.
15
DR. GARIMELLA:
16
DR. KIM:
17
DR. GARIMELLA:
Sorry.
Could you repeat, please? With respect to PPA-2
18
approaches and combination studies, do they always
19
give us complementary outcomes?
20
we do to ensure the outcomes are complementary or
21
positive like?
22
DR. KIM:
If not, what can
Combination strategies always
A Matter of Record (301) 890-4188
301
1
result in a complementary outcome.
I'm
2
paraphrasing, but the concept is adding two drugs
3
will work together in a pathway concerted effort,
4
but often do we ever see where they're antagonistic
5
toward each other?
Is that the correct paraphrase?
6
DR. GARIMELLA:
7
DR. JIN:
Yes, some inhibitor --
I think the short answer is
8
probably, at least based on my experience, no.
We
9
probably have more failed examples than successful
10
examples, especially if it's a combination of two
11
new molecular entities.
12
to both efficacy and also tolerability challenge.
13
From the efficacy perspective, many times
Some cases it really gets
14
from a mechanism-based perspective, we feel the two
15
target agents for the pathway perspective, it
16
should be complementary to each other.
17
should be at least additive or hopefully
18
synergistic efficacy results or either we have
19
great combo efficacy data from xenograft models.
20
The one panning out in the clinical situation,
21
sometimes it does not necessarily work, even from
22
early efficacy perspective.
There
Maybe it's because the
A Matter of Record (301) 890-4188
302
1
human tumor is much more heterogeneous.
2
the mutations sometimes maybe do not always match
3
with each other.
4
not very predictive, especially for the combo
5
setting, where many xenograft models are more
6
homogenous rather than heterogeneous.
7
Some of
Many of our xenograft models are
On tolerability, that's probably more cases
8
for the combo where we failed for tolerability.
9
Either the two molecules have some overlapping
10
toxicity, or sometimes actually just to our
11
surprise, there are some unexpected worsened
12
tolerability profiles when you are dosing two
13
compounds in combination.
14
back to many of the comments discussed before, the
15
lack of translatability from preclinical drugs to
16
clinical, and also the challenge of whether it's
17
target mediated or non-target mediated clinical
18
tolerability.
19
DR. GARIMELLA:
I guess this also goes
Thank you.
Is there any
20
basis on which we decide choosing the
21
[indiscernible] approach before we decide to go for
22
[indiscernible] approach?
What do we consider
A Matter of Record (301) 890-4188
303
1 2
actually as a basis for that? DR. BAILEY:
By the approach, do you mean
3
which of the two drugs to optimize during the
4
combination or where to stop with them?
5
DR. GARIMELLA:
Yes, I mean that, and is
6
there an outcome that we keep in mind before we
7
choose the approach?
8 9
DR. JIN: evolving area.
I think that probably is still an To me, the key will be how
10
to effectively do data integration and also how to
11
effectively do translation.
12
combination therapy, you want to integrate the past
13
learnings, whether in vitro, or preclinical, or
14
clinical, of each of the two molecules as a single
15
agent, to maximize the learning from there to see
16
whether you can kind of predict or have a better
17
knowledge of whether the combination may work.
18
Also, another aspect is do smart trial
Especially for
19
designs on a combination base, how to quickly
20
optimize both of the agents in a smart way, trying
21
to maximize the combo efficacy and minimize the
22
combo tolerability.
A Matter of Record (301) 890-4188
304
1
DR. PETRUZZELLI:
By the time we get to
2
clinic with a combination, all of our screening is
3
for positive results in the preclinical models.
4
When you asked about antagonism, we basically don't
5
go -- we go in there with the most promising ones,
6
and it usually is the toxicity and finding the
7
right dose.
8 9
On the other hand, I don't think we know how to really look at true synergy of the molecules.
10
We tend to look at what we need to do with one
11
molecule in a pathway and the other.
12
we can really evaluate true synergy yet in the
13
clinical setting, we're not there, in our
14
experience.
15
DR. GARIMELLA:
16
DR. JANNE:
But whether
Thank you.
Maybe I can follow up on that.
17
How are our different organizations trying to do
18
combinations in the era of immunotherapy, where you
19
have even less models to go after or to use?
20
do you do in that situation?
21 22
DR. PETRUZZELLI:
Well, we are rapidly
trying to get models.
A Matter of Record (301) 890-4188
What
305
1
(Laughter.)
2
DR. PETRUZZELLI:
There's that question.
3
think that there's a massive push for the
4
immunotherapy combinations, and people right now
5
are taking advantage of where they're seeing
6
responses with each one right now.
7
question is I think models are next.
8 9
DR. SHAW:
But the
But it's a good point because
there already are immunotherapy/targeted therapy
10
combos out there, and they're really not based on
11
preclinical --
12 13 14 15
I
DR. PETRUZZELLI:
No.
I think they're based
on where you see activity. DR. SHAW:
That's right.
There's a bit of
trial and error.
16
DR. PETRUZZELLI:
17
DR. JIN:
Absolutely, yes.
I actually feel for any new
18
scientific area
like in oncology -- any new
19
scientific area, you started with the empirical
20
approach and trial and error.
21
fishing approach, trying to combine things.
22
based on the initial results, you learn from that.
A Matter of Record (301) 890-4188
You probably do a And
306
1
And then you integrate that information, and it
2
becomes more of doing things more smartly than just
3
being trial and error.
4
process. DR. PETRUZZELLI:
5
There's probably that
I think you would agree
6
that new molecular entity and new -- it's our
7
biggest challenge right now, whether it be immuno
8
or not.
9
putting that information back into the preclinical
We're, I think, learning, again, that
10
models is really important because it hasn't been
11
easy.
12
DR. BAILEY:
Just one comment to this.
The
13
guideline from the FDA on the combinations is
14
really targeted on the combinations of two NCEs.
15
think in the NCE space, where you're combining two
16
NCEs and do a lot of preclinical data, and synergy
17
in multiple dose levels, you get some understanding
18
of how to play with both drugs.
19
I
I think the challenge for me is combining
20
with standards of care.
When you combine with a
21
standard of care, there are two options.
22
into the line of therapy where the standard of care
A Matter of Record (301) 890-4188
You go
307
1
is, and then you, in some sense, are forced to go
2
with that dose level, and that limits your ability
3
to bring the targeted agent to a level of activity,
4
or you go into a relapsed or refractory setting,
5
where the standard of care is no longer having the
6
activity that we want.
7
Combining with it there, optimizing based on
8
activity in a setting where that drug is
9
potentially no longer adding the same level of
10
activity that you would want, leads you into a
11
situation where you're probably not really truly
12
optimizing the activity of that combination for the
13
next line.
14
So you could maybe work in this setting and
15
understand that together it's active.
16
the buy-in to go up front to change the standard of
17
care dose in the line of therapy where the standard
18
of care is, it's very difficult within a company,
19
and then to be able to study outside.
20
that's the key challenge for combinations of the
21
new drugs with anything that's already out there.
22
DR. BARRETT:
But getting
And I think
One of the other things that I
A Matter of Record (301) 890-4188
308
1
think is an investment at many companies is
2
pursuing the system pharmacology models and
3
particularly for combination therapies where you
4
have the potential to have more difficult signals
5
to detect.
6
away, and you really need to vet this against good
7
preclinical and clinical data to make the virtual
8
patients that you ultimately build credible.
The problem is this is a long time
9
Again, these types of models have
10
historically been very deterministic without the
11
incorporation of appropriate levels of variability
12
uncertainty.
13
future.
14
can bear fruit, but it's going to require feedback
15
and for us to do targeted experiments to explain
16
some of the uncertainty in it.
17
They're an investment, really, in the
But I think it's something that over time
That's where I would come back to our
18
discussion earlier today, is when you take a look
19
at these plots of all these overlapping
20
exposures -- I mean, we keep asking for more PK
21
data, but how much is going to be helpful here when
22
you can't discriminate the doses or you have very
A Matter of Record (301) 890-4188
309
1
unclear exposure-response relationships?
2
need to do something a little different than we're
3
doing it.
4
DR. KIM:
What is that, though?
5
question.
6
forward with innnovative ideas?
So we
That's the
Any suggestions where we can move
7
DR. BARRETT:
8
(Laughter.)
9
DR. JIN:
Jin?
Just ask my --
10
(Laughter.)
11
DR. KIM:
Because I definitely think there's
12
an inordinate amount of translation from
13
nonclinical to the clinical trial setting.
14
what translation are we doing, and how can we move
15
forward from translating the clinical trial results
16
to the real-world population?
17
there, but it's so hard to get a hold of it and to
18
filter out the signal.
19
But
The data's out
Then, really, the postmarketing data, as we
20
venture into trying to analyze claims outcome data
21
and make any sense of anything like that, or even
22
using electronic health records systems, it's very
A Matter of Record (301) 890-4188
310
1
difficult to get a hold of actually what's going on
2
with the dose relationship, with the adverse event,
3
with concomitant or preexisting illnesses that
4
patients not in clinical trial carry, and how that
5
affects because, as Amy alluded to, we don't study
6
those patients.
7
more healthy trial population.
8
introduce these drugs with the polypharmacy that is
9
really American health care, how do these drugs
We really have a very homogenous, And when we
10
really interact with those drugs, and what is the
11
functional consequences to the actual patient?
12
That's a question that we need to really start
13
addressing. DR. BARRETT:
14
You talked about this at the
15
beginning, about communication.
This is one area
16
where access to the patient population, the payer
17
community, they're engaged right now in other
18
therapeutic areas.
19
sponsors in designing clinical -- those real-world
20
trials.
21
oncology, but it's coming.
22
diabetes space, that's absolutely the case.
They're involved with the
So that maybe hasn't trickled down to I can tell you in the
A Matter of Record (301) 890-4188
311
Again, the electronic health records, I
1 2
think everybody's -- I'm going to use the word "big
3
data" because everybody's using that as their
4
current buzzword, but it's not clear exactly what
5
we're going to find yet.
6
trying to develop advanced analytics, but they're
7
mostly predictive models without understanding. Not that we shouldn't do it because I think
8 9
And I think everybody's
it's a movement in the right direction, but those
10
systems were not built to capture more granular
11
data.
12
practice.
13
It doesn't mean it's the optimal practice.
14
it's uniquely tied to
15
bit disconnected from what we really would need.
16
They just happened to capture the current It doesn't mean it's the best practice.
DR. KIM:
Yes.
And
billing, so it's a little
I'm also looking forward
17
to -- as a result of this, that we could start
18
moving more.
19
systems pharmacology and some of this more
20
analytical approach to the data that we have.
21
ongoing discussions really on the ground floor is
22
really important to kind of understand where we can
I think we heard some illusion to
A Matter of Record (301) 890-4188
So
312
1
move forward with this and the potentials that we
2
can use.
3
DR. DONOGHUE:
I hope it's not off topic,
4
but I wanted to talk a little bit -- Martha
5
Donahue, FDA.
6
training.
7
feedback a little bit about whether you think a lot
8
of concepts that we've discussed over the past few
9
days are applicable to pediatric oncology trials.
10
I'm a pediatric oncologist by
And I just wanted to get the panel's
As you know, we usually have -- I don't know
11
if it's a luxury or not, but we usually have the
12
adult MTD before a lot of these agents go into
13
pediatrics.
14
MTD, and generally the trials I see are 3 plus 3 or
15
Rolling 6 designs to identify a dose for children.
16
So I was curious as to whether you
We start at 80 percent of the adult
17
think -- to my mind, a lot of the comments I've
18
heard discussed are even more relative to
19
pediatrics due to the small patient population and
20
the need for increased efficiency to find the
21
appropriate doses and drugs for these rare patient
22
subsets that have druggable, targetable diseases.
A Matter of Record (301) 890-4188
313
1
But I also know there are a lot of barriers. So I was just curious to see what your
2 3
thoughts might be on the relevancy to drug
4
development of pediatric patients at this time,
5
whether you think there are any unique barriers to
6
applying these concepts to pediatric oncology
7
trials.
8
think that we at the agency could do to help smooth
9
out some of the barriers.
10
DR. YATES:
11
concepts couldn't be used.
12
mathematical modeling to integrate data, things
13
like once you've got a more mechanistic
14
understanding of the pharmacokinetics with the
15
physiologically based pharmacokinetic modeling, you
16
can take into account the differences and maturing
17
physiology in children to understand how you need
18
to change the dose to get the same exposure.
19
And finally, if there are things that you
It's a tall order.
I can't think of why the The idea of using
As adults, once you've got prior information
20
of tolerability in the adults, that gives you an
21
idea we'r on the dose curve in pediatrics as well.
22
I guess the challenges are not really being
A Matter of Record (301) 890-4188
314
1
involved pediatric development myself, but the
2
challenges I have seen is in terms of confirming
3
pharmacokinetics in children.
4
sample as intensively in children as you can in
5
adults.
6
DR. BARRETT:
Obviously, you can't
Even with the benefit of
7
having the adult MTD defined, I think these trials
8
are very hard in pediatrics, predominantly because
9
you even have a smaller population, difficult to
10
identify, difficult to get them enrolled.
11
the things I fundamentally struggle with, with
12
those trials, is that they take an outrageous
13
amount of time to complete.
14
One of
I think, especially, we have to have a more
15
innovative way of picking the winner and stopping
16
trials with the likelihood that we're going find
17
something because it's just not ethical to keep
18
those children on a trial when you know you're not
19
going to find something.
20
compounds in development, and I think a more
21
innovative way of looking at a pick-the-winner
22
design is what's needed here.
You have a lot of
A Matter of Record (301) 890-4188
315
DR. PETRUZZELLI:
1
If we have a strong
2
compelling reason, genetic reason or whatever, we
3
will enter pediatrics early, and we apply the same
4
rules.
5
investigators.
6
situation where you need to streamline it and
7
really make it focused.
8
can be done.
9
And they have not been as slow as I thought they
10
And we have very strong partnerships with I do think you need to be in a
But it can be done, and it
And we apply a lot of the same rules.
were going to be.
11
I agree with the pick the winner.
We have
12
these complicated diseases, and we think there's
13
more than one target, and we want to go in
14
combination, I thin, that's an important strategy,
15
to look at more than one combination, and genetics
16
can drive that oftentimes.
17
DR. BAILEY:
I'm from the same company.
18
We've got numerous examples of phase 1 designs that
19
are using the methodology we presented, where
20
rather than taking preclinical data as the
21
historical data, we're taking adult data.
22
always try 18 percent of the adult MTD.
A Matter of Record (301) 890-4188
We don't We may
316
1
even start a pediatric study before we've actually
2
got the adult MTD; close to.
3
We've seen activity.
We've got the good
4
hypothesis Lilli mentioned from the genetic side to
5
go into a specific disease.
6
just relying purely on the pediatric data -- that's
7
a primary goal -- but we're actually looking at
8
similarity or dissimilarity between what we're
9
seeing in the pediatric patients and the adult
10
patients, helping to inform where to go in the
11
pediatric side.
12
But we then are not
The dosing may not be a flat dosing in the
13
pediatric settings, also looking at the exposure
14
relationships and translating that information.
15
But accounting for uncertainty about how it
16
translates directly -- because you can't just take
17
an adult exposure and translate it into a pediatric
18
patient, so you have to incorporate some of that
19
with an increased uncertainty about that.
20
But they're actively ongoing.
And as Lilli
21
said, well designed settings don't take necessarily
22
as long as you think.
And by using that comparison
A Matter of Record (301) 890-4188
317
1
to the adult, to the previous point, you are able
2
to start looking at picking winners earlier.
3
DR. DONOGHUE:
Thank you.
4
DR. DE ALWIS:
Dinesh De Alwis, Merck.
5
Actually, looking at the crizotinib example, which
6
was very eloquently presented by Dr. Alice Shaw, I
7
was just wondering, given the fact that we're
8
venting as a MET inhibitor -- it's the phase 1, and
9
it was pure luck, serendipity, whatever you want to
10
call it.
11
with these responses, and we found that it was down
12
to ALK.
13
And the science came at the same time
I'm just wondering, if you go back to the
14
graveyard of drugs that have failed about 10 years
15
ago, 5 years ago, what work are we doing to look
16
at -- base on the science we know not in terms of
17
repurposing, understanding -- there were some
18
significantly successful drug that we've actually
19
just shelved away.
20
outcomes, I think we should be doing more.
21 22
For the sake of patient
DR. PETRUZZELLI: good point.
I think that's a really
We have one that we've actually just,
A Matter of Record (301) 890-4188
318
1
I guess resurrected is the right term from when we
2
paused it.
3
there was enough question in the balances, and then
4
data emerged.
5
it's potentially something we can move forward
6
with.
7
We didn't know where to use it.
And
And we took a look at it again.
DR. SHAW:
And
I would say with crizotinib,
8
you're actually right about the timing because had
9
the study started in 2005 instead of 2006 -- I
10
mean, there was really no signs of activity early
11
on in that dose escalation that I showed you.
12
mean, these were not molecularly selected patients
13
in dose escalation.
14
discovered around that time, who knows how much
15
interest the sponsor would have had just focusing
16
on MED or rare ALK-positive lymphomas, so that
17
crizotinib may not have come to bear had the timing
18
not been perfect.
19
DR. DE ALWIS:
20
DR. KIM:
I
And had there not been ALK
Thank you.
At this time, I'd like to open up
21
the floor for any final comments, then I'm going to
22
open up the panel for any final comments, too.
A Matter of Record (301) 890-4188
But
319
1
if anybody has anything pressing that they want to
2
say for final take-home messages, things that we
3
should learn, things that the agency maybe needs
4
some feedback on, we'll welcome those, too, as
5
well.
6
anybody on the panel wants to finalize any
7
comments, now is your chance.
8 9
But if anybody wants to open the floor or if
DR. JIN:
Actually, I have a question
probably for the panel that just came to my mind.
10
We talk about shared learnings, and also in the
11
past two days, we see a lot of successful case
12
examples, where basically it's compounds that's
13
approved.
14
probably more failed examples than positive
15
examples, and there may be at least, personally I
16
feel, more effective learnings from the failed
17
examples:
18
efficacy or it's tolerability.
19
I'm sure for every company, there are
why those failed, whether it's a lack of
Is there anything we can learn from those or
20
any channels for us to learn from the failed
21
examples?
22
panel or in the audience wants to comment on that.
I'm just wondering if everyone on the
A Matter of Record (301) 890-4188
320
1
DR. PETRUZZELLI:
2
DR. BARRETT:
That's a good point.
I will add something.
I
3
believe you're right.
I think part of the issue
4
for every company dealing with this is our data
5
streams are focused on success.
6
building towards a submission activity assuming
7
success, so it's difficult to capture with
8
appropriate granularity the failures.
9
have that habit of doing it.
Everything is
We don't
There is a lot more efforts in the earlier
10 11
stages I think in accumulating databases, so our
12
preclinical data tends to be more
13
regard.
14
that -- particularly if you're talking about doing
15
things like clinical trial simulation, this is a
16
problem.
17
because our priors are built on successes and not
18
failures.
19
should all invest in more.
20
DR. JANNE:
rich in that
But I think it is something
We tend to have rosy-colored glasses
So this is I think an issue that we
I would say the failed
21
combinations don't also make it into the literature
22
as often as the successful ones do, so you may not
A Matter of Record (301) 890-4188
321
1
always know why they failed or even learn from
2
that.
3
learnings in there, and then should make an effort
4
to get them published. DR. PETRUZZELLI:
5 6 7
But I do agree that there are definitely
I think it's a great
point. DR. JIN:
Because also, I think for
8
patients -- because ultimately, we want to benefit
9
the patients.
We don't want the patient to suffer
10
also from multiple tryings just by different
11
sponsors of maybe a similar targeted combination.
12
It's just people don't know and are open sharing
13
that information.
14
DR. McKEE:
I want to make one comment about
15
breakthrough therapy designation because I've heard
16
it over and over in the past two days that a
17
company receives breakthrough therapy designation,
18
and then they're all of a sudden in a rush to get
19
their application in.
20
That's not really what breakthrough therapy
21
designation is.
What it is, is we see a clinical
22
signal that you may have a very good product.
A Matter of Record (301) 890-4188
And
322
1
because it's such a new program, we've had a wide
2
variation in where products are when they receive
3
their breakthrough therapy designation.
4
some where you've had very few patients for the
5
clinical signal, and we've had some where,
6
basically, the trial got supported approval.
7
randomized trial was what supported the
8
breakthrough therapy designation.
9
We've had
A
So I want to make clear that, first of all,
10
it doesn't guarantee approval and also the
11
opposite.
12
designation, or you don't think you have a
13
breakthrough therapy product, that does not mean
14
it's not going to be a successful product.
15
If you don't get breakthrough therapy
So I think the sort of rush to get your
16
product on the market once you have a breakthrough
17
therapy may be hindering some good development, and
18
that really it should be used as a tool to
19
facilitate conversations about how to best develop
20
it and not how to most quickly develop it.
21
know that it may run against some commercial
22
interest, but I think maybe sometimes you've had
A Matter of Record (301) 890-4188
And I
323
1
some breakthrough therapy products where the rush
2
to develop it and get it on the market may not have
3
always led to the best development program. DR. RATAIN:
4
Can I just add something to
5
that?
It's my perspective, someone who works
6
neither for the FDA nor for industry, that there
7
are too many drugs in development, and that the
8
competition among companies is leading to emphasis
9
on speed rather than quality of development.
And
10
I'm just wondering what others think about that.
11
And if that is a problem, what the industry is
12
doing about that problem.
13
DR. DAMBACH:
14
absolutely a consideration.
15
may be harsh, but it's a business.
16
confronted with needing a billion dollars just to
17
develop the 90 percent of your losses because we
18
still really haven't figured out how to improve our
19
success.
20
our committee meetings.
21 22
Well, I would say that it's And you know -- this When you're
These are real discussions that happen at
It's really this balance between how do we move as quickly as possible, especially when
A Matter of Record (301) 890-4188
324
1
there's competition, which there always is, and put
2
together the best product that we can?
3
the real discussions.
4
we're not going to pretend that it's not a
5
consideration, especially with breakthrough.
6
You're absolutely right.
7
word, we're told to go fast.
8
(Laughter.)
9
DR. DAMBACH:
Those are
We try to balance those, but
As soon as we hear that
That is exactly it, and we're
10
told to set up scenarios of success and mitigation,
11
but that's exactly what that means, quite honestly.
12
DR. McKEE:
To respond to that, I'm not
13
trying to say that the agency won't take a risk in
14
an accelerated approval in particular if you have
15
breakthrough therapy designation because we don't
16
want a slow development either.
17
certainly since breakthrough therapy designation
18
has come up, I think we've moved very quickly to
19
approve some products in a very quick fashion,
20
faster than we probably would have before, even
21
though we traditionally have been fairly quick to
22
make our decisions within the oncology division.
A Matter of Record (301) 890-4188
And I think
325
1 2 3
So again, we're willing to take that risk if you truly have a breakthrough therapy product. DR. DAMBACH:
And I hear you.
I mean, I
4
appreciate your position, too.
5
meaningful thing to keep reiterating.
6
comment was that probably the first thing on the
7
company's mind is to try to move quickly when they
8
hear that.
9 10
DR. BAILEY:
I think it's a I guess my
I think it's not just for the
company that receives the breakthrough status.
11
(Laughter.)
12
DR. DAMBACH:
13
DR. BAILEY:
Yes. When one drug receives
14
breakthrough status, it's any company who's got a
15
backup -- not a backup but another company.
16
is certainly thrown into this whirlwind of we have
17
to go quickly.
18
because we've got to get this in as quickly because
19
otherwise, we're not going to be the first one.
20
That
Just abandon all hope and go
I think that to the companies -- and clearly
21
for the regulators when reviewing these, you still
22
have this level of quality that you expect on the
A Matter of Record (301) 890-4188
326
1
drug and the level of information on the drug.
2
that, unfortunately, isn't consistent on all
3
approvals anyway.
4
individually, and internally we suddenly change our
5
mind on what's important.
6
problem, just the interpretation of breakthrough.
7
DR. SHAW:
And
Every single drug is looked at
I think that is a real
I think from the clinical end, we
8
of course have no financial incentive for rushing
9
it, but our rush really is based on patient need.
10
And it is very true that there are so many patients
11
out there, EGFR patients, outpatients, who cannot
12
access clinical trials to get these drugs.
13
too sick.
14
financial resources.
15
these drugs hit the market, there are patients who
16
are just barely hanging on, and they finally get
17
these drugs.
18
they live a lot longer.
19
to new opportunities as well.
20
care standpoint, it makes a huge difference.
21 22
They can't travel.
They're
They don't have the
And all of us have seen when
And they totally turn around, and
DR. PETRUZZELLI:
And they now have access So from a patient
We felt that the whole way
through.
A Matter of Record (301) 890-4188
327
1
DR. McKEE:
One other regulatory point I
2
would like to make, Geoff dispelled one myth about
3
what we think about 3 plus 3.
4
another myth.
5
within a label for different diseases, for
6
different patient populations within a same
7
disease.
8
So we're open to that when you have the data to
9
support it.
Well, I'll dispel
We're not opposed to multiple doses
We've talked about brain mets versus non.
I think it's perfectly acceptable to
10
have different doses in the label for different
11
populations.
12
DR. ROY:
Amit Roy from BMS.
I just
13
actually wanted to make a very similar point that
14
Dr. Shaw just made.
15
group that actually isn't represented here is the
16
patient advocates.
17
presumption that going into an approval with
18
suboptimal dose, that everyone knows it's
19
suboptimal but actually is benefiting a group of
20
patients is not a very good thing.
21
good thing temporarily before the dose optimization
22
is done.
A very important stakeholder
There's the underlying
It could be a
So I think in having that conversation,
A Matter of Record (301) 890-4188
328
1
we ought to have the patient advocates also present
2
to get their input as well, what they think. DR. JIN:
3
I have to admit, I'm also in
4
conflict myself during the sessions because on one
5
hand, we want to make sure we want to collect the
6
best data set as possible, do as much dose
7
optimization, getting the right drug, at the right
8
dose, to the right patient.
9
best-ever product.
We want to have that
However, especially for the breakthrough
10 11
situation, actually, there is a very good reason,
12
especially for the patient's benefit, to run fast
13
with a very lean as possible package.
14
certain perspectives, it's probably good to get
15
something to these patients to save their life, to
16
some of the previous points also made before,
17
rather that something, best possible, but after six
18
months.
Because to
Sometimes I've heard the comment of
19 20
regulators are requiring dose optimization, PMC,
21
PMR.
22
like a negative image.
Maybe it's not -- because many times it's It's like, oh, you didn't
A Matter of Record (301) 890-4188
329
1
get the dose right in the first place.
2
times, the other way to think about it, if it is a
3
conscious risk, it's a conscious decision, it's a
4
risk waiting to take.
5
But many
Maybe from another perspective for some
6
cases, it may be a smart risk to get because you go
7
over with some dose to the patient, get it approved
8
so the patient starts using it, and then you can do
9
further refinement after the dosing, to dose on
10
additional trials, and potentially maybe even have
11
additional amendment or revising of the label,
12
either adding additional dose ranges in there or
13
even potentially changing that.
14
Of course, I know there are operational,
15
logistical, and commercial concerns regarding
16
changing doses in a label, but also, some of
17
these -- I think especially the breakthrough
18
mechanism, to have this efficacy brings benefit to
19
the patients quickly.
20
optimization topic, of course we want to do as much
21
and smartly during drug development.
22
cases, maybe it is okay to dose optimization after
Maybe for the dose
A Matter of Record (301) 890-4188
But for some
330
1
the approval, and that may be many times a smart
2
risk to take. Adjournment
3
C O N T E N T S
4 5
AGENDA ITEM
PAGE
6
Welcome and Work Objectives
7
Amy McKee, MD
5
8
Pasi Janne, MD, PhD
5
9 10 11 12 13
SESSION 3: Dose-Exposure Exploration Dose-Finding Studies Qi Liu, PhD
6
Dose Optimization: Ceritinib Dan Howard, PhD
14
Dose Optimization: Axitinib as a Case
15
Example for Dose Titration
16
Yazdi Pithavala, PhD
17
Dose Optimization: Vandetanib
18
Eric Masson, PharmD
19 20 21 22
22
46
72
Dose Optimization: Combinations Jin Jin, PhD
72
Moderated Panel Discussion Qi Kiu, PhD
A Matter of Record (301) 890-4188
125
331
1 2
Julie Bullock, PharmD Audience Q&A
323
3
C O N T E N T S (continued)
4 5
AGENDA ITEM
6
SESSION 4: Integrating Dose Optimization in
7
Clinical Development
8
Integrative Approach to Dose Finding
9
PAGE
Geoffrey Kim, MD
10
Feasibility of an Integrative, Adaptive
11
Dose-Finding Trial
12
Rajeshwari Sridhara, PhD
13
Barriers to Implementing Integrative,
14
Adaptive Dose-Finding Trials
15
Lilli Petruzzelli, PhD
16
Practical Considerations of Implementing
17
Dose Optimization Strategies in the Clinic
18
Alice Shaw, MD, PhD
19
Approaches to Integrative, Adaptive
20
Dose-Finding Combination Studies
21
Pasi Janne, MD, PhD
22
A Matter of Record (301) 890-4188
6
22
46
72
72
332
1 2 3
C O N T E N T S (continued)
4 5
AGENDA ITEM
PAGE
6
Future Considerations and Moving Forward
7
Geoffrey Kim, MD
8
Lilli Petruzzelli, PhD
9 10
125
Audience Q&A
323
Wrap Up and Adjourn
380
11 12 13 14 15 16 17 18 19 20 21 22
A Matter of Record (301) 890-4188
333
Adjournment
1
DR. KIM:
2
I think the story was a great
3
example of that where approval was there, but the
4
sponsor took on the responsibility to find the
5
better dose for the patients.
6
are in that same boat, and I think we want to
7
explore that because we are in the breakthrough era
8
as well, that as we accelerate the drug development
9
to give to the patients, we really need to start
And I think we all
10
looking at this as a life cycle, a life span
11
development, how do we continuously improve on the
12
metrics that we have. Any further comments or poignant points
13 14
while you have the floor?
15
(No response.)
16
DR. KIM:
Otherwise -- I really want to
17
thank everybody again on behalf of
18
our -- especially, I want to thank our co-sponsor,
19
AACR, who's been a wonderful host.
20
getting five disciplines to really participate on
21
the same clinical trial without any fist fights is
22
great.
A Matter of Record (301) 890-4188
And really,
334
1
(Laughter.)
2
DR. KIM:
But we really appreciate your
3
participation, and I think we've all learned so
4
much, a tremendous amount.
5
build on this and continue to engage in these
6
critical clinical trial issues, as well as trying
7
to translate this directly to the patient benefit.
8
So thank you all for participating.
9 10 11
And I think we can
(Applause.) (Whereupon, at 4:05 p.m., the meeting was adjourned.)
12 13 14 15 16 17 18 19 20 21 22
A Matter of Record (301) 890-4188
View more...
Comments
