WEBVTT

NOTE
This file was generated by Descript 

00:00:00.000 --> 00:00:03.019
Judith: Welcome to Berry's In the
Interim podcast, where we explore the

00:00:03.019 --> 00:00:06.900
cutting edge of innovative clinical
trial design for the pharmaceutical and

00:00:06.900 --> 00:00:09.210
medical industries, and so much more.

00:00:10.130 --> 00:00:10.869
Let's dive in.

00:00:12.240 --> 00:00:16.060
Well, welcome everybody
back to In The Interim.

00:00:16.510 --> 00:00:17.970
I'm your host, Scott Berry.

00:00:20.560 --> 00:00:21.340
All right.

00:00:21.640 --> 00:00:27.830
We cover a number of different
topics on In The Interim, uh, w- some

00:00:27.830 --> 00:00:29.770
recent sports editions and all that.

00:00:30.250 --> 00:00:36.750
Today I'm gonna go back a little bit, and
I'm gonna go back to a really interesting

00:00:36.750 --> 00:00:43.790
effort of multiple clinical trials coming
together for questions during COVID.

00:00:44.830 --> 00:00:49.270
This effort, uh, was given a
name called the Multi-Platform

00:00:49.270 --> 00:00:51.160
Randomized Clinical Trial.

00:00:51.800 --> 00:00:53.280
Now, I'm gonna come back to the name.

00:00:53.280 --> 00:00:56.460
It's a, it's a really important
part of the story actually.

00:00:57.080 --> 00:01:00.820
But let me set up the story, and I
think it will make a little bit more

00:01:00.820 --> 00:01:05.510
sense as to why that name is, is such
an important part, uh, of the story.

00:01:07.600 --> 00:01:08.100
Okay.

00:01:08.800 --> 00:01:15.240
A great deal of people, time, and
effort went into this, so I'm gonna

00:01:15.240 --> 00:01:18.220
give some, some explanation of this.

00:01:18.220 --> 00:01:22.710
I'm gonna walk through a number of
people from here at Berry Consultants,

00:01:23.050 --> 00:01:28.930
Roger Lewis, Lindsey Berry, uh, Liz
Lorenzi, Mark Fitzgerald, Michelle Detry,

00:01:28.930 --> 00:01:31.360
Anna McLaughlin, Christina Saunders.

00:01:32.140 --> 00:01:35.390
We have huge numbers of people
globally working on this.

00:01:35.390 --> 00:01:41.420
I can't possibly mention everyone, um,
uh, in this, but a, a huge effort and

00:01:41.420 --> 00:01:43.460
I, and I think you'll, you'll hear why.

00:01:44.890 --> 00:01:45.290
Okay.

00:01:46.370 --> 00:01:48.860
So there were...

00:01:49.520 --> 00:01:55.380
A long time ago in a place far, far
away, there were, there were three

00:01:55.380 --> 00:02:00.370
different trials, and I'll, I'll, I'll
try to set them up briefly and then

00:02:00.370 --> 00:02:02.520
describe the intersection of them.

00:02:03.340 --> 00:02:04.060
So plat...

00:02:04.060 --> 00:02:09.050
Yeah, you, you, you're probably aware
that platform trials played this enormous

00:02:09.050 --> 00:02:16.650
role during COVID for finding therapies
that did and did not work for the

00:02:16.650 --> 00:02:19.930
treatment of patients with COVID-19.

00:02:21.340 --> 00:02:26.040
Vaccines were a whole different issue,
but for therapeutics to treat COVID,

00:02:26.710 --> 00:02:32.960
the Recovery trial, the Principal
trial, the REMAP-CAP trial, the, the

00:02:33.350 --> 00:02:42.380
active programs, the US NIH, the US,
um, uh, Operation Warp Speed efforts.

00:02:42.880 --> 00:02:49.170
There was ICE by COVID, the Together
trial, Solidarity, DNDi, ATTACK.

00:02:49.170 --> 00:02:52.790
So multiple platform trials
made huge impacts during it.

00:02:53.700 --> 00:02:56.130
So I'm gonna describe three
of them somewhat briefly.

00:02:56.130 --> 00:03:01.930
REMAP-CAP, which you're gonna hear
if you tune into this, uh, uh,

00:03:02.020 --> 00:03:04.590
podcast for, for more interims.

00:03:04.940 --> 00:03:07.020
You're gonna hear much
more about this effort.

00:03:07.470 --> 00:03:13.430
Amazingly, we are on episode
sixty-two Two or three, and I've

00:03:13.430 --> 00:03:15.520
not talked about Remap-Cap yet.

00:03:15.720 --> 00:03:17.660
It's, it's an incredible effort.

00:03:17.660 --> 00:03:23.640
In part, it's just such a, um,
uh, enor-- it was such an, an

00:03:23.690 --> 00:03:26.760
enormous and very cool effort.

00:03:26.760 --> 00:03:30.090
I, I haven't quite figured out
how to do podcasts about it.

00:03:30.570 --> 00:03:32.510
So this is, uh, Remap-Cap.

00:03:32.510 --> 00:03:38.140
Remap stands for Randomized Embedded
Multifactorial Adaptive Platform.

00:03:39.270 --> 00:03:42.750
It was built in 2015.

00:03:44.400 --> 00:03:48.330
It was built for a potential pandemic.

00:03:48.920 --> 00:03:56.760
Yes, 2015, and you know COVID-19
was largely a 2020, uh, uh, disease.

00:03:57.250 --> 00:04:06.690
So Remap-Cap was built in 2015 and started
enrolling patients in multiple countries

00:04:07.340 --> 00:04:09.860
that had community-acquired pneumonia.

00:04:10.070 --> 00:04:15.710
The idea was that if there's a
pandemic, it's very likely it's going

00:04:15.710 --> 00:04:17.910
to go through intensive care units.

00:04:18.210 --> 00:04:22.360
It's going to be something that looks
like community-acquired pneumonia,

00:04:22.800 --> 00:04:28.360
and the trial started enrolling with
multiple therapies that are very

00:04:28.360 --> 00:04:32.840
interesting in and of themselves in
non-pandemic community-acquired pneumonia.

00:04:33.860 --> 00:04:40.830
And a master protocol was built, and
it built something called a sleeping

00:04:40.830 --> 00:04:47.590
strata that was-- We called it the
pandemic strata, and it could adopt

00:04:47.590 --> 00:04:49.650
and say, "We now have a pandemic.

00:04:49.650 --> 00:04:53.880
We are enrolling that pandemic,
and we already have therapies.

00:04:53.880 --> 00:04:55.180
We already have a database.

00:04:55.180 --> 00:04:56.800
We already have sites enrolling.

00:04:57.160 --> 00:04:59.670
This is a platform trial ready to go."

00:04:59.790 --> 00:05:01.930
And in fact, that's exactly what happened.

00:05:03.440 --> 00:05:12.490
So Remap-Cap adopted the sleeping
pandemic strata in February of 2020, very,

00:05:12.490 --> 00:05:16.460
very early, and one of the domains...

00:05:16.460 --> 00:05:18.610
Well, uh, uh, uh, let me
go backwards a little bit.

00:05:19.720 --> 00:05:24.060
Uh, Remap-Cap from the beginning
adopted two stratifications.

00:05:24.370 --> 00:05:28.280
Now, stratifications in that trial
are really, really, uh, different

00:05:28.280 --> 00:05:29.980
than what you m- might be used to.

00:05:30.690 --> 00:05:36.470
That, that implies that therapies
are modeled f- prospectively,

00:05:36.470 --> 00:05:42.610
that they may have differential
efficacy in those groups, and it

00:05:42.610 --> 00:05:45.340
was severe state and moderate state.

00:05:46.450 --> 00:05:49.070
All of these patients are hospitalized.

00:05:50.220 --> 00:05:54.810
Moderate state are hospitalized but
are con-- are not considered severe.

00:05:54.810 --> 00:05:56.870
It's sort of the complement of severe.

00:05:57.200 --> 00:06:03.880
Severe state is that you're
hospitalized, and you have ICU-level

00:06:04.295 --> 00:06:08.775
Organ support, cardiovascular
or respiratory organ support.

00:06:09.305 --> 00:06:19.155
Largely ventilator, vasopressors, uh,
ECMO, uh, uh, ICU-level organ support.

00:06:19.195 --> 00:06:20.785
You are severe state.

00:06:21.045 --> 00:06:25.745
Moderate state, you're hospitalized
in the ward, but you don't have

00:06:25.885 --> 00:06:29.135
severe-level, um, um, organ support.

00:06:30.055 --> 00:06:36.765
The primary endpoint is for the
pandemic strata is organ support free

00:06:36.765 --> 00:06:44.085
days, and that's through 21 days, and
it actually had a neat aspect of it.

00:06:44.085 --> 00:06:49.635
So this is an ordinal endpoint where
mortality is the worst outcome.

00:06:50.435 --> 00:06:55.935
We, we refer to that as a minus one,
and that was actually 90-day mortality.

00:06:55.965 --> 00:07:01.485
If a patient died through 90 days,
they are considered a minus one.

00:07:01.705 --> 00:07:02.785
Actually, I, I wanna be...

00:07:02.835 --> 00:07:05.185
I wanna, uh, make sure I'm clear on that.

00:07:05.495 --> 00:07:12.325
It's in-hospital mortality, and that,
that extends through 90 days of exposure.

00:07:12.915 --> 00:07:17.845
So if the patient dies in their hospital
visit, they are considered a minus one.

00:07:18.695 --> 00:07:23.855
And then it's ordinally the days
they are free of organ support, given

00:07:23.855 --> 00:07:25.765
they survive and leave the hospital.

00:07:26.585 --> 00:07:29.825
So zero is the second-worst outcome.

00:07:30.145 --> 00:07:34.065
It means you survive and left
the hospital, but for 21 days you

00:07:34.065 --> 00:07:35.765
were not free of organ support.

00:07:36.135 --> 00:07:40.965
And then one, two, three, all
the way up to 21 days, you were

00:07:40.965 --> 00:07:43.975
free of organ support, which
can't happen in the severe state.

00:07:43.975 --> 00:07:45.175
You start on organ support.

00:07:47.125 --> 00:07:51.405
There, there were many domains
that were started in this, and

00:07:51.405 --> 00:07:54.205
this is a multifactorial platform.

00:07:54.955 --> 00:08:02.075
They were enrolling antibiotics, steroids,
uh, from the beginning in, in CAP.

00:08:02.075 --> 00:08:09.275
As soon as the pandemic came, they
started investigating steroids, macrolide,

00:08:09.655 --> 00:08:16.995
antivirals, immune modulation, uh,
interferon, Anakinra, tocilizumab,

00:08:16.995 --> 00:08:20.645
sarilumab, convalescent plasma.

00:08:21.175 --> 00:08:27.815
And then they adopted very early
in March a domain that was looking

00:08:27.815 --> 00:08:30.885
at therapeutic anticoagulation.

00:08:32.335 --> 00:08:39.215
And the notion here is that the,
the cytokine storm at the time was

00:08:39.525 --> 00:08:46.725
discussed as the, the, the coagulation
was part of that within the body

00:08:47.045 --> 00:08:52.240
that was causing the severe disease
of COVID Now I'm a statistician, so

00:08:53.270 --> 00:08:57.680
please don't take this as medical, but
that's largely in my statistical mind

00:08:57.680 --> 00:09:03.020
what was going on, and the idea is
that would therapeutic anticoagulation

00:09:03.020 --> 00:09:06.310
heparin be beneficial for patients?

00:09:06.310 --> 00:09:10.100
Would it improve their survival
in getting off of organ support?

00:09:10.890 --> 00:09:17.000
So a domain was adopted with two arms,
standard dose thromboprophylaxis,

00:09:17.000 --> 00:09:19.390
so, uh, uh, the, uh...

00:09:19.690 --> 00:09:26.030
Consider this low dose, which is
given standardly or therapeutic

00:09:26.030 --> 00:09:29.090
dose, a high dose of anticoagulation.

00:09:29.280 --> 00:09:30.600
Is that beneficial?

00:09:31.110 --> 00:09:35.190
And so it started randomizing
those two options.

00:09:36.370 --> 00:09:41.990
Patients and doctors were not blinded
to which arm they were in, uh, uh,

00:09:42.050 --> 00:09:47.250
within this setting, but that was
adopted by the, the REMAP-CAP trial.

00:09:47.250 --> 00:09:49.980
It started enrolling, sorry,
in April of twenty-twenty.

00:09:52.200 --> 00:10:01.630
ATTACK is a, uh, Canadian-funded, and
at the time, a Canadian-funded platform.

00:10:01.910 --> 00:10:06.610
The initial goal of this was to
investigate exactly the same question,

00:10:06.640 --> 00:10:11.710
therapeutic anticoagulation versus
prophylactic dose anticoagulation.

00:10:12.500 --> 00:10:17.170
Enrolling moderate and severe in
Canada, actually, this extended a

00:10:17.170 --> 00:10:20.190
little bit into the US, South America.

00:10:20.540 --> 00:10:24.390
So this was a platform for
investigating the exact same question.

00:10:25.020 --> 00:10:31.280
They also, interestingly, in moderate
disease, they stratified by D-dimer level.

00:10:32.270 --> 00:10:40.840
D-dimer is a biomarker that i-it has
some indication of high levels of fibrin

00:10:41.390 --> 00:10:49.210
in the blood, es-essentially suggesting
coagulation is going on, intravascular

00:10:49.210 --> 00:10:53.580
coagulation, and the notion is this
thought that this might measure,

00:10:54.130 --> 00:10:59.030
again, as a statistician, levels of
coagulation, the cyto-cytokine storm.

00:10:59.320 --> 00:11:03.640
These may be patients that particularly
have differential effect due to

00:11:04.610 --> 00:11:06.770
therapeutic dose anticoagulation.

00:11:07.430 --> 00:11:13.390
So their design stratified
severe disease and then within

00:11:13.390 --> 00:11:16.310
moderate, high and low D-dimer.

00:11:16.410 --> 00:11:20.710
So really three different groups the
trial could come out with a conclusion.

00:11:21.300 --> 00:11:26.100
Originally, their endpoint was
a twenty-eight-day, are you

00:11:26.100 --> 00:11:28.050
alive and free of organ support?

00:11:29.460 --> 00:11:32.960
As we're gonna see, they start
to work together, and they adopt

00:11:33.410 --> 00:11:34.960
the same endpoint as REMAP-CAP.

00:11:37.020 --> 00:11:39.430
ACTIV-4-A was one of the...

00:11:39.430 --> 00:11:45.940
A third platform now, was a, is
an NIH-funded trial investigating

00:11:47.175 --> 00:11:50.715
It ended up investigating multiple
things, but the first thing to

00:11:50.715 --> 00:11:56.005
investigate is therapeutic dose heparin
versus prophylactic dose heparin.

00:11:56.715 --> 00:12:00.495
There are multiple investigators
involved in these trials that are

00:12:00.495 --> 00:12:03.625
involved in, in, in the same trials.

00:12:05.515 --> 00:12:11.515
Uh, me personally, I was a statistician
for each of these three trials.

00:12:12.315 --> 00:12:15.755
So I had been working on REMAP-CAP
since two thousand and fifteen.

00:12:15.795 --> 00:12:18.265
I still work on REMAP-CAP today.

00:12:19.275 --> 00:12:24.365
Uh, ATTACK, we had worked with the
Canadian investigators, uh, uh,

00:12:24.395 --> 00:12:30.845
Patrick Lawler, Ryan Zarichansky,
and, um, and many others in that.

00:12:31.285 --> 00:12:33.745
And we were working with
them on their design.

00:12:33.745 --> 00:12:37.085
And then when ACTIVE4 came around,
we worked with multiple other

00:12:37.085 --> 00:12:40.575
statisticians on the design of this.

00:12:40.575 --> 00:12:43.205
And that was the first
question for them, therapeutic

00:12:43.205 --> 00:12:45.905
anticoagulation versus prophylactic.

00:12:46.415 --> 00:12:48.285
Again, hospitalized patients.

00:12:49.435 --> 00:12:52.585
Organ support-free days
is the primary endpoint.

00:12:53.395 --> 00:12:58.985
And now this is somewhat of the third to
be adopted, and at this point, there's

00:12:58.985 --> 00:13:03.625
already discussions about these three
p- platforms doing very similar things.

00:13:04.185 --> 00:13:08.055
So they adopt organ support-free
days as the primary endpoint.

00:13:09.365 --> 00:13:13.245
They are going to stratify
differential conclusions by

00:13:13.245 --> 00:13:19.135
moderate and severe, and also within
moderate by the two D-dimer levels.

00:13:21.565 --> 00:13:22.625
They adopt...

00:13:22.905 --> 00:13:27.165
By the way, all three of these
trials adopt a Bayesian approach.

00:13:28.265 --> 00:13:32.135
And maybe that was correlated to me
being involved in these, but they

00:13:32.135 --> 00:13:34.515
all three adopt a Bayesian approach.

00:13:35.025 --> 00:13:40.285
So the que-- the, the, the landscape here
is this is globally considered a very

00:13:40.285 --> 00:13:49.055
important question, and there are these
three trials that are all separately

00:13:49.105 --> 00:13:51.705
addressing the exact same question.

00:13:52.545 --> 00:13:57.875
So what happens when one of them
reaches a conclusion, maybe the other

00:13:57.875 --> 00:14:03.635
one doesn't reach that same conclusion,
but maybe it has po- similar data and

00:14:04.015 --> 00:14:06.095
papers come out at different times?

00:14:06.705 --> 00:14:09.965
It was thought to be, uh, somewhat messy.

00:14:10.345 --> 00:14:15.805
Okay, we can do a meta-analysis
of the three trials.

00:14:16.595 --> 00:14:18.575
They don't carry the same weight.

00:14:18.975 --> 00:14:22.175
Their perception of them
is somewhat different.

00:14:22.735 --> 00:14:26.295
And so the three trials are now
talking and figuring out, "What

00:14:26.295 --> 00:14:27.995
do we do in this scenario?"

00:14:28.755 --> 00:14:35.445
The most important part of it is each
of the three trials working separately

00:14:36.095 --> 00:14:40.805
are going to take longer to come
up with an answer to this question

00:14:41.115 --> 00:14:43.775
in the middle of a raging pandemic.

00:14:44.315 --> 00:14:48.680
Twenty twenty Where the pandemic
is globally raging, and this is

00:14:48.680 --> 00:14:51.950
an important question, all three
are gonna take longer separately.

00:14:53.020 --> 00:14:57.430
So the investigators, the funding
groups, everybody comes together and

00:14:57.430 --> 00:15:00.560
say, "We're gonna work together, and
we're gonna actually do it in a way.

00:15:00.560 --> 00:15:05.310
We're gonna create something brand
new, and we're gonna create something

00:15:05.310 --> 00:15:09.670
new, and it's gotta be differentiated
from a meta-analysis because it is

00:15:09.670 --> 00:15:11.070
different than a meta-analysis."

00:15:11.580 --> 00:15:14.900
Statistically, it has components
of that, but that this is...

00:15:14.900 --> 00:15:20.440
We're gonna call it a multi-platform
randomized clinical trial.

00:15:20.480 --> 00:15:22.340
It's a randomized clinical trial.

00:15:23.360 --> 00:15:30.630
The three efforts all decide to pool their
data together into a single analysis.

00:15:32.000 --> 00:15:35.850
They're gonna create a joint
analysis plan where they all

00:15:35.850 --> 00:15:38.430
sign on board prospectively.

00:15:39.140 --> 00:15:44.000
They will not read out separately, so it's
not that you're gonna see a publication

00:15:44.000 --> 00:15:46.940
from REMAP-CAP and then the MPRCT.

00:15:47.540 --> 00:15:52.570
They all agree they'll publish
together, they'll com- they'll

00:15:52.570 --> 00:15:56.490
do their adaptations together,
they'll combine their data together.

00:15:56.490 --> 00:15:58.450
This is a randomized clinical trial.

00:15:58.960 --> 00:16:04.600
It's just that these three trials all take
their data from all of their global sites.

00:16:04.940 --> 00:16:10.110
They put it together prospectively
in a, in a combined analysis.

00:16:11.100 --> 00:16:14.930
Hence, we called it the
multi-platform, rather obviously

00:16:14.930 --> 00:16:16.640
the three platforms coming together.

00:16:16.670 --> 00:16:20.580
It's a randomized clinical trial, and
we wanted to make sure that message got

00:16:20.580 --> 00:16:24.250
across that this is not a meta-analysis.

00:16:24.250 --> 00:16:28.380
This is the prospective primary
analysis of all three trials.

00:16:29.950 --> 00:16:34.760
There was quite an operational effort
to this, as you can imagine, that all

00:16:34.760 --> 00:16:42.260
three of these trials combine their
outcome data together into a, an, an

00:16:42.260 --> 00:16:44.660
unblinded statistical analysis committee.

00:16:45.820 --> 00:16:49.530
That committee carries
out the primary analyses.

00:16:50.700 --> 00:16:56.620
They create a efficacy report, and then
they talk to the three different DSMBs.

00:16:56.620 --> 00:17:01.600
Each of these platforms have a DSMB,
and they did that simultaneously.

00:17:02.360 --> 00:17:09.200
So these three DSMBs all come together
and meet together because they're

00:17:09.200 --> 00:17:15.060
hearing the same analyses of the
combined data of the three trials.

00:17:17.680 --> 00:17:19.840
Okay, so an incredible effort.

00:17:19.910 --> 00:17:24.560
There is a statistical analysis plan
agreed to by the three platforms.

00:17:25.140 --> 00:17:29.670
Uh, uh, that's finalized
on August 29th, 2020.

00:17:31.060 --> 00:17:37.650
And, uh, ACTIVE starts enrolling in,
I, I, I think it was largely, uh...

00:17:37.690 --> 00:17:38.790
Do I have that written down?

00:17:38.790 --> 00:17:45.430
ACTIVE, uh, starts enrolling patients,
I believe summer, June-ish Within that.

00:17:45.950 --> 00:17:50.560
So all the trials I have started
enrolling, no analyses have been done.

00:17:50.810 --> 00:17:53.570
August twenty-ninth, the
analysis plan is finalized.

00:17:55.090 --> 00:17:58.260
They're gonna do monthly
a-adaptive analyses.

00:17:59.420 --> 00:18:05.090
The adaptive analyses which could trigger
in severe disease is its own group.

00:18:06.450 --> 00:18:11.580
And then in moderate, there's
low and high D-dimer levels.

00:18:11.610 --> 00:18:15.800
So there are three inferential groups
where we could reach differential

00:18:15.800 --> 00:18:18.180
occlusions, uh, conclusions or the same.

00:18:19.760 --> 00:18:22.750
What happened was early on in
the trial, they ended up with a

00:18:22.750 --> 00:18:27.000
large amount of missing D-dimer
data for moderate patients.

00:18:27.440 --> 00:18:30.330
So we had moderate patients
with low, high, and there

00:18:30.330 --> 00:18:31.720
was a group that was missing.

00:18:31.720 --> 00:18:34.410
We created our own group and
said it's missing D-dimer.

00:18:35.560 --> 00:18:40.040
But it's really awkward to come up
with a conclusion, so we, we reported

00:18:40.040 --> 00:18:44.960
on that strata, and it was part of
the modeling, but it couldn't trigger

00:18:44.960 --> 00:18:48.230
separately because that would be
really awkward to say, "Well, if you're

00:18:48.230 --> 00:18:50.320
missing D-dimer, here's what we think."

00:18:50.780 --> 00:18:54.420
So those three other groups
could have adaptive conclusions.

00:18:55.020 --> 00:19:00.510
Could conclude superiority if the
probability that the odds ratio

00:19:01.450 --> 00:19:07.540
for proportional odds ratio model,
Bayesian model for organ support-free

00:19:07.540 --> 00:19:13.410
days, this combination of mortality
and days free of organ support, if

00:19:13.410 --> 00:19:17.510
the probability of superiority is
greater than ninety-nine percent.

00:19:18.420 --> 00:19:19.980
We're gonna do monthly analyses.

00:19:19.980 --> 00:19:23.640
If it's above ninety-nine percent,
we're gonna call that a trigger.

00:19:24.800 --> 00:19:30.060
We're gonna say it's futile if
the probability is greater than

00:19:30.060 --> 00:19:33.880
ninety-five percent that the odds
ratio is less than one point two.

00:19:34.780 --> 00:19:40.430
That was considered a reasonable
clinical threshold where we're gonna

00:19:40.430 --> 00:19:44.040
say the effect is a-at best small.

00:19:45.740 --> 00:19:51.330
And harm is a ninety-nine percent chance
that it has an odds ratio less than one.

00:19:53.499 --> 00:19:54.839
So that's the design.

00:19:54.989 --> 00:19:59.049
Single Bayesian analysis combining
all these patients together.

00:19:59.559 --> 00:20:07.759
A critical part of the primary analysis
is a Bayesian hierarchical model that

00:20:07.869 --> 00:20:15.589
allows dynamic borrowing, and the model
was really a model that borrows moderate

00:20:16.189 --> 00:20:21.809
patients, low and high, and missing
D-dimer are part of a hierarchical model.

00:20:23.339 --> 00:20:28.979
The mean of that model is in a
hierarchical model with the severe effect.

00:20:29.759 --> 00:20:34.989
So it's this two-tiered hierarchical model
where the moderate D-dimer levels can

00:20:34.989 --> 00:20:39.919
borrow from each other because they're
all within the same disease state of

00:20:39.919 --> 00:20:45.960
moderate And then the moderate and severe
effects can borrow if they're similar.

00:20:48.640 --> 00:20:49.160
Okay.

00:20:50.010 --> 00:20:57.220
Covariates adjusted in the model, regions,
sites, age, critically important, time.

00:20:57.880 --> 00:21:00.220
Uh, there are no
non-concurrent controls here.

00:21:00.700 --> 00:21:05.370
Uh, all randomized concurrent
controls within them, but we adjust

00:21:05.370 --> 00:21:08.820
for time because it's such an
important thing within the pandemic.

00:21:09.140 --> 00:21:16.950
It-it-- Both for disease variations,
but also because of, of, uh, times of

00:21:16.950 --> 00:21:20.050
surge where maybe outcomes are different.

00:21:21.230 --> 00:21:21.700
Okay.

00:21:23.950 --> 00:21:30.620
In the first adaptive analysis,
November 20th, twenty-twenty, no...

00:21:30.660 --> 00:21:32.540
I-I'm blinded to all of this.

00:21:32.980 --> 00:21:34.390
Uh, no triggers are met.

00:21:34.470 --> 00:21:35.920
We're told to continue.

00:21:37.430 --> 00:21:40.990
Um, the DSMBs all meet,
no triggers are met.

00:21:42.470 --> 00:21:47.560
On December nineteenth, six days
before Christmas, twenty-twenty,

00:21:47.960 --> 00:21:54.050
adaptive analysis number two occurs,
and a severe state triggers hit.

00:21:54.440 --> 00:21:58.720
Within the hierarchical model, there are
twelve hundred and seven patients are

00:21:58.720 --> 00:22:06.760
met, and therapeutic anticoagulation meets
futility, meaning ninety-five percent

00:22:06.760 --> 00:22:12.680
chance or higher that the effect has
an odds ratio less than one point two.

00:22:13.060 --> 00:22:18.000
Those results are publicly disclosed,
but the-- no paper comes out.

00:22:18.270 --> 00:22:24.520
Uh, randomization to severe is stopped
in all three platforms, and publicly

00:22:24.520 --> 00:22:27.060
disclose this, this futility trigger.

00:22:28.670 --> 00:22:33.590
So patients immediately in all
three of these global platform

00:22:33.590 --> 00:22:37.650
trials stop randomizing to
therapeutic anticoagulation.

00:22:40.090 --> 00:22:45.060
Now, the, the, the message is
continue in the moderate state.

00:22:46.360 --> 00:22:50.590
So all three platforms are still
enrolling in the moderate state.

00:22:51.260 --> 00:22:52.720
We don't know the answer yet.

00:22:54.660 --> 00:23:01.460
Adaptive analysis number
three happens January 22 2021

00:23:04.310 --> 00:23:10.340
At that analysis, superiority
for therapeutic anticoagulation

00:23:10.490 --> 00:23:16.670
in the moderate state is met for
both high and low D-dimer groups.

00:23:17.480 --> 00:23:25.490
2,200 patients, just over 2,200 patients
go into that analysis, and superiority

00:23:25.490 --> 00:23:32.650
greater than a 99 chance therapeutic
anticoagulation is beneficial in

00:23:32.650 --> 00:23:37.685
patients in the moderate state And
it's the same conclusion for D-dimer.

00:23:38.235 --> 00:23:39.315
vary by D-dimer.

00:23:39.835 --> 00:23:43.455
Randomization is stopped in
each of the three platforms.

00:23:43.455 --> 00:23:45.725
Those, uh, results are announced.

00:23:47.455 --> 00:23:52.145
The, the data aren't disclosed to
the level of typical publications.

00:23:52.145 --> 00:23:53.125
That comes out.

00:23:54.975 --> 00:23:57.985
So I, I, I, I'll, I'll come back to that.

00:23:58.015 --> 00:24:04.415
So two papers are published in the New
England Journal of Medicine, and they

00:24:04.415 --> 00:24:09.685
are published side by side, and they come
out in the summer of twenty twenty-one.

00:24:11.755 --> 00:24:13.335
And you can, uh...

00:24:13.515 --> 00:24:14.405
easy to find.

00:24:14.405 --> 00:24:19.715
Therapeutic anticoagulation with
heparin in critically ill, that's

00:24:19.715 --> 00:24:23.725
what they refer to the steer--
severe state, patients with COVID-19.

00:24:24.305 --> 00:24:28.915
And then therapeutic anticoagulation
with heparin in non-critically

00:24:28.915 --> 00:24:30.625
ill patients with COVID-19.

00:24:31.425 --> 00:24:36.155
Back-to-back papers in the New England
Journal of Medicine report out on

00:24:36.155 --> 00:24:42.255
this, and the fascinating thing
about that is the data are the same.

00:24:42.445 --> 00:24:48.385
One model is run, and the results are
presented in two different papers.

00:24:49.895 --> 00:24:54.425
And, um, the result of each is
done separately, yet the Bayesian

00:24:54.425 --> 00:24:56.065
hierarchical model is shrinking.

00:24:56.625 --> 00:24:59.725
Now, they're shrinking within
the moderate state, potentially,

00:24:59.725 --> 00:25:01.645
and across moderate and severe.

00:25:02.225 --> 00:25:06.845
By the differential conclusion, you
can guess, of course, that one was

00:25:06.845 --> 00:25:09.215
futility and one was superiority.

00:25:09.605 --> 00:25:15.915
The model, uh, learned that the effect
was differential and didn't borrow much.

00:25:16.465 --> 00:25:21.455
But yet the component that borrowed
in the moderate state between D-dimer

00:25:21.455 --> 00:25:27.505
levels did shrink those values together,
enabling the conclusion to happen that,

00:25:28.245 --> 00:25:32.645
uh, uh, therapeutic anticoagulation
is beneficial in the moderate state.

00:25:35.915 --> 00:25:41.615
Now, these conclusions, th-this effort
coming together, the whole idea of this,

00:25:42.395 --> 00:25:47.615
if they would not have come together,
it's very likely these conclusions

00:25:47.615 --> 00:25:52.995
would have been months later and may
have been differential because they

00:25:52.995 --> 00:25:59.425
didn't combine them together, leading
to even more conclusi-- more confusion.

00:26:00.295 --> 00:26:01.695
So they all come together.

00:26:01.695 --> 00:26:02.575
They read out.

00:26:02.575 --> 00:26:05.635
The patients are combined together.

00:26:05.815 --> 00:26:13.325
It's done faster and more effectively
by the three groups doing this together.

00:26:14.875 --> 00:26:19.915
Now, it was an incredibly
incredible operational exercise

00:26:20.245 --> 00:26:22.385
to have this done all together.

00:26:22.885 --> 00:26:27.145
Huge amount of time and effort done
to have this happen, but it made

00:26:27.145 --> 00:26:33.240
a huge impact in the disease Now,
in the moderate state, the final

00:26:33.240 --> 00:26:37.800
primary analysis comes out with a
ninety-eight point six probability

00:26:38.530 --> 00:26:43.840
that therapeutic dose anticoagulation
is superior in the overall group, and

00:26:43.840 --> 00:26:49.930
it's ninety-seven and, uh, ninety-three
and ninety-seven in the three groups.

00:26:49.930 --> 00:26:55.280
In the Bayesian hierarchical model,
the estimate o-overall is about a one

00:26:55.280 --> 00:27:01.050
point two seven, one point three one,
one point two two, one point three two.

00:27:01.740 --> 00:27:08.510
So one point two seven is the,
uh, uh, moderate state estimate

00:27:08.840 --> 00:27:12.280
with the bottom of the c-credible
interval being one point zero three.

00:27:13.090 --> 00:27:17.900
So statistically significant
with that ninety-eight point six

00:27:17.900 --> 00:27:21.820
probability, it's beneficial for
patients in the moderate state.

00:27:24.120 --> 00:27:32.630
Now, in the severe state, the probability
is ninety-five percent that it is harmful.

00:27:35.380 --> 00:27:39.430
That-- And the odds ratio
is point eight three.

00:27:39.430 --> 00:27:44.850
We set up the odds ratio, so an odds ratio
less than one were negative outcomes.

00:27:45.290 --> 00:27:50.190
Increased mortality and organ
support through-throughout the scale.

00:27:50.990 --> 00:27:56.710
Ninety-nine point nine percent probability
it was futile, less than one point two.

00:27:57.060 --> 00:28:01.010
The top of the ninety-five percent
credible interval is one point zero three,

00:28:01.450 --> 00:28:07.590
exactly where the moderate state bottom
of the interval was, one point zero three.

00:28:09.890 --> 00:28:13.660
Ninety-nine point nine probability
of futility, ninety-five

00:28:13.660 --> 00:28:18.650
percent probability of harm of
therapeutic dose anticoagulation.

00:28:20.370 --> 00:28:26.530
So this effort was unbelievable in
coming together, the impact of this.

00:28:26.850 --> 00:28:31.810
The result is an amazing thing
of these trials together.

00:28:32.290 --> 00:28:33.220
And let me sort of...

00:28:33.450 --> 00:28:37.750
The, the, the role of borrowing
within these analyses and the

00:28:37.750 --> 00:28:43.070
role of prospectively identifying,
these were not post hoc subgroups.

00:28:43.930 --> 00:28:49.980
These were the analysis plan started in
2015 actually, in the Remap-Cap trial

00:28:50.280 --> 00:28:52.730
and adopted by all for this domain.

00:28:54.380 --> 00:28:59.700
If you would have pooled the data
and said, "We want one conclusion

00:28:59.700 --> 00:29:05.760
in COVID-19," the odds ratio,
pooling those groups together,

00:29:05.760 --> 00:29:11.670
adjusting for all of the states and
covariates, would have been 1.03

00:29:12.640 --> 00:29:17.230
Essentially 1 with the
bottom of the interval 0.85

00:29:17.250 --> 00:29:17.720
the top 1.22

00:29:19.420 --> 00:29:24.800
It would have reached a conclusion of
futility, uh, a-and and no difference.

00:29:24.800 --> 00:29:29.755
Therapeutic anticoagulation doesn't
matter If you would have pooled them

00:29:29.755 --> 00:29:31.705
all together in these two groups.

00:29:32.415 --> 00:29:35.815
Remember the conclusion through
the Bayesian hierarchical

00:29:35.815 --> 00:29:45.555
model was harm, likely harm,
in severe, benefit in moderate.

00:29:47.235 --> 00:29:49.765
Guidelines adopted these results.

00:29:49.855 --> 00:29:54.135
There's a lot of communication to
guideline results that these weren't post

00:29:54.135 --> 00:29:57.315
hoc, prospectively set up in the SAP.

00:29:58.155 --> 00:30:02.905
Because of that, therapeutic
anticoagulation is given to

00:30:02.905 --> 00:30:04.295
patients in moderate state.

00:30:04.425 --> 00:30:05.965
It is not in severe.

00:30:07.155 --> 00:30:12.925
Unclear what happens if they
don't do this differential

00:30:12.925 --> 00:30:19.225
analysis, hierarchical modeling,
uh, of these results within this.

00:30:21.295 --> 00:30:21.665
Okay.

00:30:21.665 --> 00:30:27.085
Uh, I, I-- and again, go to
these papers and look them up.

00:30:27.255 --> 00:30:31.005
Uh, uh, it was amazing effort within that.

00:30:31.895 --> 00:30:38.825
The, the, the, the take-home of this
is a brand new entity, multi-platform

00:30:38.825 --> 00:30:39.905
randomized clinical trials.

00:30:39.905 --> 00:30:42.585
We are talking about this in
a number of other efforts.

00:30:42.585 --> 00:30:50.455
So this continues to go on that there's
multiple separate funded platforms that

00:30:50.455 --> 00:30:53.245
are investigating the same question.

00:30:53.845 --> 00:30:59.745
Rather than them, the two of them compete,
be the first one out, let's combine our

00:30:59.745 --> 00:31:02.195
data together and have one conclusion.

00:31:02.525 --> 00:31:08.055
It's a more heterogeneous patient
population, bigger data, faster results.

00:31:08.705 --> 00:31:13.025
During this effort in COVID, it was
an amazing effort, amazing effort.

00:31:14.105 --> 00:31:20.855
Differential HTE, the Bayesian modeling
completely changes the result around.

00:31:21.115 --> 00:31:27.125
Benefit and harm rather than it doesn't
matter, which had been the answer

00:31:27.125 --> 00:31:30.035
if that was not done in the trials.

00:31:34.785 --> 00:31:35.725
All right.

00:31:37.035 --> 00:31:42.285
I hope you enjoyed this
look back into COVID.

00:31:42.705 --> 00:31:47.655
Um, we-- lots of people
jumper- jumping in in effort.

00:31:48.005 --> 00:31:52.965
The, the scientific, uh, aspects of
this are very much worth revisiting,

00:31:52.965 --> 00:31:59.105
even though COVID is, is not
nearly the, the, um, uh, medical

00:31:59.105 --> 00:32:01.285
challenge that it was at the time.

00:32:01.715 --> 00:32:07.415
Many scientific lessons, platform
trials, amazingly successful.

00:32:07.705 --> 00:32:14.725
And here is a new entity called the
multi-platform randomized clinical trial.

00:32:17.495 --> 00:32:20.025
You heard about those interim analyses.

00:32:20.765 --> 00:32:24.065
We stay here in the interim.

00:32:24.665 --> 00:32:25.825
Thanks for joining.

00:32:26.035 --> 00:32:27.155
Until next time