REMAP-CAP Results: Oseltamivir in Critically-Ill Influenza Patients

In this episode of "In the Interim…", Dr. Scott Berry speaks with Dr. Srinivas Murthy, Dr. Thomas Hills, and Dr. Lindsay Berry about the REMAP-CAP trial results on oseltamivir in critically ill influenza patients. The trial used a Bayesian covariate-adjusted platform design and found oseltamivir was not effective at reducing 90-day mortality with a “98% and 99% probability of harm in 90-day mortality” compared to control. Covariate adjustment addressed baseline and site variation. Subgroup analyses showed greater harm in patients with higher illness severity. Sensitivity analyses using alternative neutral, optimistic, and pessimistic priors produced important scientific exploration of the results. No evidence was found for benefit over control in any subgroup.

The discussion highlights the first randomized, controlled evidence in this patient group, contrasting prior observational studies and clinical guidelines. A mechanism of harm remains unclear. REMAP-CAP is continuing enrollment in moderate severity and pediatric cohorts to further examine population-specific effects. The episode also addresses the broader challenges of trial design and interpretation in acute care research, the limitations of nonrandomized evidence, and the importance of ongoing Bayesian analyses and transparent reporting.

Key Highlights

• Pre-print is available: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7172531
• REMAP-CAP platform trial, Bayesian logistic regression, covariate adjustment
• Oseltamivir arms: statistical trigger for inferiority, “98 and 99% probability of harm”
• Greater harm in sicker subgroups, consistent results across sensitivity analyses
• Ongoing arms: moderate severity and pediatric cohorts, mechanistic questions unresolved
• Context: limitations of previous historical data studies, clinical practice impact, future research directions

For more, visit us at https://www.berryconsultants.com/

Creators and Guests

Host

Scott Berry

President and a Senior Statistical Scientist at Berry Consultants, LLC

What is In the Interim...?

A podcast on statistical science and clinical trials.

Explore the intricacies of Bayesian statistics and adaptive clinical trials. Uncover methods that push beyond conventional paradigms, ushering in data-driven insights that enhance trial outcomes while ensuring safety and efficacy. Join us as we dive into complex medical challenges and regulatory landscapes, offering innovative solutions tailored for pharma pioneers. Featuring expertise from industry leaders, each episode is crafted to provide clarity, foster debate, and challenge mainstream perspectives, ensuring you remain at the forefront of clinical trial excellence.

Judith: Welcome to Berry's In the
Interim podcast, where we explore the

cutting edge of innovative clinical
trial design for the pharmaceutical and

medical industries, and so much more.

Let's dive in.

Scott: Welcome everybody
back to In the Interim.

I'm your host, Scott Berry.

So today we have a special episode.

We have another episode on the results
of a clinical trial, so we are gonna

talk about the results of a domain,
uh, intervention on the REMAP-CAP

trial, and this episode is coming out
concordant with a preprint of the results.

And so we will put a link on the
website where you can get a preprint.

We're, we're hoping you have that
as we walk through this result.

I think you'll find this an interesting
clinical trial, an interesting

result, uh, and, and exciting science.

Okay, so, uh, I have guests
today to talk to these results.

I have the chair of the International
Trial Steering Committee for

REMAP-CAP, Dr Srinivas Murthy.

Uh, Srin is an infectious disease
specialist and a clinical associate

professor, Department of Pediatrics,
Faculty of Medicine, University

of British Columbia, so we have,
uh, Canadian representation.

We also have Dr Thomas Hills.

Tom is a clinical immunologist and
infectious disease physician at Te

Toka Tū Māu Auckland Hospital, and the
infectious disease program lead at the

Medical Research Institute of New Zealand.

And as, as Tom was just saying as
we were starting, the flu season is

starting in the Southern Hemisphere,
so we have representation there.

He was the 2024 Sir Charles Hercus
Fellowship of the Health Research

Council of New Zealand, uh,
recipient, and was a Rhodes Scholar.

Interesting.

And we have Dr Lindsey Berry.

She is, uh, Lindsey is a senior
statistical scientist at Berry

Consults, and this is maybe her third
or fourth time on In the Interim.

is a, uh, co-lead of the analysis
team for REMAP-CAP and also

a member of the design team.

There's also an unblinded statistical
analysis committee that sees all

of the REMAP-CAP unblinded data.

We, on the design team and the analysis
team, only see the results once a trigger

has been met and the results become
public and the analysis takes place.

And, uh, by the way, huge thanks to those
on the statistical analysis committee.

They do tremendous work.

Uh, Roger Lewis, Anna McLoughlin,
Michele Detry, Mark Fitzgerald,

Christina Saunders, uh, on that
committee, they do a wonderful job.

So welcome everybody to In the Interim

Lindsay Berry: Thank you.

Happy to be here

Scott: Okay, so let's,
let's talk about this trial.

So this is oseltamivir for the
treatment of severe influenza, we're,

we're hoping you have this result.

Srin, you talked about this at
Critical Care, Care Reviews.

You can get a copy of that, uh, uh, if you
go to the Critical Care Reviews, uh, uh,

website, you can see that presentation.

So that's the general one.

We'll get to the results, but Srin, why
don't you tell us about the patients

this is going to be relevant for?

srinivas murthy: Yeah.

And so what we're looking at
is critically ill patients with

confirmed influenza infection.

And so influenza, as many listeners
know, it's a circulating virus,

happens in seasonal disease
primarily, causes severe respiratory

infections, makes you critically ill.

and so what we are looking at is
that population, those patients who

end up in an ICU of their influenza
infection that's been confirmed

by some microbiological test.

And the reason we're looking at those
patients is that there's not really

any treatments available for them.

Um, there's a lot of supportive
care interventions that we do,

but that population doesn't
have any targeted therapies that

have been found to be useful

Scott: Okay.

So, um, that's the populations,
that's, that's the patients.

And what, what, what does a typical
patient like this look like?

Sort of general age ranges of
this and, and severity of disease?

srinivas murthy: Yeah.

And so in the critically, critically ill
population with flu, the median age is

about 60, which goes along with a lot
of different critical illness syndromes.

Although influenza affects all
ages, and so there's patients

downwards to small children, there's
adolescents, there's the very elderly.

Um, but just like any and most other
critical illness syndromes, the 60 to

70-year-old is the burden of disease
in the ICUs that's the highest.

Um, these patients are very sick.

Lots of them require invasive ventilation.

Um, if not invasive ventilation, then
non-invasive ventilation, so that's

positive pressure masks or lots of
oxygen via high-flow nasal cannula.

And so pretty sick patients who all
have a probability of mortality at

least 10% to 20% depending on the
cohort that we look at across the world

Scott: Okay, so we are gonna talk
about the use of oseltamivir for,

uh, and, and what happened with
oseltamivir, and hopefully people

have the results here and, and so
you've seen the results of this.

The, the treatment, the
intervention, what is oseltamivir?

Um

Tom Hills: I can take
that question, Scott.

Yeah, o- oseltamivir
is not a new medicine.

It's, it's been around a long time,
it's an oral medicine, so you can

take it at home, in the hospital.

It can be given to critically
ill patients, and it blocks the

replication of the influenza virus.

So it's an antiviral, and it's
targeted at the influenza virus.

Blocks an influenza, uh,
enzyme called neuraminidase.

Uh, and so oseltamivir's been around
a long time, it's been studied,

uh, in many different contexts.

There are randomized trials for
the prevention of influenza.

Somebody coughs next
to you in the hospital

Scott: a four-metered room,

Tom Hills: then

Scott: you

Tom Hills: you

Scott: might diagnose something
that isn't appropriate

Tom Hills: the development of an
influenza illness and becoming

unwell, like Shrin's alluded to.

Or you might take

Scott: because I think it did,

Tom Hills: uh, very early

Scott: early in the illness at

Tom Hills: home

Scott: home drive me

Tom Hills: duration of symptoms.

And there's lots of
data in those contexts.

But there's no data in critically ill
patients from randomized clinical trials.

over the years, what

Scott: what we've known about
ourselves coming in, how we view

Tom Hills: more data has accrued,

Scott: ourselves after the exchange.

Uh, so as more and more data is improved

Tom Hills: become clear that
oseltamivir is quite effective

at preventing influenza.

If you're exposed to influenza and you
take oseltamivir you can roughly sort

of halve your chances of developing
symptomatic influenza There was a

Scott: There's a lot of interest

Tom Hills: in using oseltamivir
very early in ill- in the illness.

Uh, if you're an outpatient at,
at home and you're trying to

avoid admission to, to hospital.

and over time, what's become clearer is
that oseltamivir is, is a potent influenza

antiviral, and that's why it works for,
for, uh, post-exposure prophylaxis.

But its clinical effect in those
patients taking it at home very

early in their illness is detectable.

Uh, it reduces the duration of symptoms
by eighteen hours in the latest

meta-analysis published in The Lancet.

because there's thousands of patients in
those studies, the confidence intervals

are quite tight, and they fall below
the twenty-four-hour threshold that

the WHO said is clinically important.

So oseltamivir, based on the evidence we
have in outpatients, definitely reduces

the duration of symptoms, less than
twenty-four hours, which was considered

the clinically important, for making
a recommendation that it is used in

those WHO, uh, influenza guidelines.

And so we've actually shifted away from
using oseltamivir in outpatients, and

it's now no longer recommended because
we have not seen data that it reduces

things like hospitalization or death in
that lower-risk outpatient community.

It does shorten the duration of symptoms,
but not by a clinically important amount.

left us with a situation, uh,
where oseltamivir is now no longer

used in the community, where it
clearly does shorten the duration

of symptoms, just not by much.

and we have patients coming to hospital
unwell, Srin described, where we

really want an effective therapy,
and it's that population in hospital

and particularly in ICU, where we
don't have clinical trial data.

And we've been left with a situation
that we don't have data, but those

are patients we really wanna help,
and oseltamivir is widely used.

It's recommended in the guidelines
despite the lack of evidence.

There is some observational
evidence supporting its use.

And clinicians are looking at sick
patients who they want to treat,

and oseltamivir is a familiar,
well-understood medication.

And so it's commonly given to people in
the, in the hospital and particularly

sicker people in the intensive care unit

Scott: All right.

Fascinating.

So, uh, a-and, and coming back to our
experiment, I mean, it's fascinating

that it's not commonly used where we
have randomized evidence it's beneficial.

It's used quite a bit where we have
no randomized, uh, uh, evidence.

We'll come back to generally in
severe disease perhaps even, as, as it

seems like there's a lack of benefit
across a, a range of, of viruses.

But our experiment was even to the point
where we have, in the trial, we're gonna

talk a lot about being randomized to an
intervention where you get no antiviral,

and we're gonna call that the control.

So that's one of the
interventions in the trial.

for five days, and I think that's,
uh, 75 milligrams twice daily

for five days, is one of the
interventions in the arms in the trial.

Another intervention is Oseltamivir
given for 10 days i-in the trial.

There are sites in the REMAP-CAP global
trial that did not want a control arm.

In severe disease that they thought that
given the guidelines that they, they,

they weren't going to give a no control.

In s-those sites, they can opt out
of the control, and they can be

randomizing between Oseltamivir five
days and 10 days, which was still, uh,

an important question to be answered.

Do you get benefit from five,
or five days or 10 days?

Are they different in the treatment?

is evidence that this is rather it was
even that people wouldn't-- uh, sites

wouldn't randomize to the control arm

Tom Hills: Yeah, that's correct, Scott.

And oseltamivir was so widely used that
in many ways you can think of the five

days oseltamivir as the arm that's most
reflective of what usually happens.

then the 10-day arm existed because
there are some guidelines and,

uh, some hospital protocols that
say in critically ill patients you

can extend treatment to 10 days.

So that was something that
was already happening.

Uh, and in many ways, the experiment
here is randomizing away from the

receipt of oseltamivir, which is given
to most patients as part of usual care.

We had a quick look at some Southern
Hemisphere data from Australian

intensive care units, and it's sort
of 80% or upwards of that, uh, of

critically ill flu patients who get
oseltamivir as part of routine care.

Um, so it's really

Scott: embedded in practice.

Tom Hills: And

Scott: You're right to identify that, um,

Tom Hills: for

Scott: for many sides,

Tom Hills: the experiment

Scott: or the randomized way of most of

Tom Hills: if

Scott: the things

Tom Hills: had equipoise
to the no antiviral control

Scott: Okay, so setting up the result.

Before the, the trigger of this
result we're gonna talk about, were,

uh, uh, multiple arms in this domain
where patients can be randomized, and

oseltamivir five days becomes the sort
of, uh, therapy that all sites have.

And so it's the mathematical
connection across all of these.

Um, and so that where control, you
might randomize patients away from

oseltamivir, and control becomes almost
an intervention, um, uh, within it.

So you could be randomized to no
antiviral, five days, oseltamivir

10 days, baloxavir, an alternative
antiviral, or a combination of baloxavir

and oseltamivir, five and 10 days.

un- that six different interventions
you can get randomized to, result that

we're gonna talk about was a trigger
for oseltamivir five and 10 days.

We are still blinded to baloxavir
or baloxavir plus oseltamivir.

We don't know the results of those.

So we're gonna focus on those
three interventions is all

we know about at this time.

Okay, and then we'll tell you what, what--
how the experiment sort of moved forward.

Okay.

So, um, the primary outcome in this
domain, what is, uh, Shrin, what's

the primary outcome for this domain?

srinivas murthy: It's a good
old-fashioned 90-day mortality

as a dichotomous primary outcome.

So very easy to interpret
for a lot of audiences

Scott: How do we analyze that, Lindsay?

How, what's the primary analysis
method for 90-day mortality?

Lindsay Berry: So the entirety of
REMAP-CAP, we use one overarching

Bayesian logistic regression model.

Um, so all the patients randomized in
REMAP-CAP are analyzed in this model, um,

including the patients in this domain.

Um, we include covariate
adjustments in the model, um, in

addition to the treatment effect.

So we adjust for important predictors
like patient's age, um, sex of the

patient, location, which is, um, the trial
site, um, which we nest within country.

We also adjust for, um,
time period of enrollment.

Uh, we have about, uh,
quarterly time periods.

So from the current time period and
then moving back all the way to the, the

beginning of REMAP-CAP when enrollment
started, we adjust for the fact that

the mortality rate could vary over time.

Um, and then we have other
important predictors, including

the severity of illness.

So whether the patient was in shock
at baseline, whether they were on

an invasive mechanical ventilation,
um, or if they had severe hypoxemia.

Um, and I mentioned this is the
same model used for all domains.

So we do adjust for whether a patient
was randomized into another domain.

So we're talking about one,
which is the influenza antiviral.

If a patient had another
assignment in another domain,

we would also adjust for that.

Um, and then we would see
only the intervention effects

to which we're unblinded.

So there are some parameters in
the model that we don't, we don't

see until they're reported out

Scott: Just to be clear for, for
people, we, enrolled-- REMAP-CAP

enrolled a large number of pandemic
COVID-19 pandemic patients.

They're not in this model.

This model is the non-COVID pandemic,
uh, as we talk about going back.

But we, we've been enrolling for
years this cohort of influenza and

non-influenza CAP, uh, for that.

So Srin, take us through the result now

srinivas murthy: Sure.

Um, and so we got a letter from our DSMB
in March of this year telling us that the

five-day and ten-day arms triggered or
hit a statistical trigger for inferiority,

um, and to stop those interventions.

Um, and what that means is we had
a posterior probability that those

arms were the best in the domain of
less than zero point two percent.

And I think when we got that letter,
I think many of us were surprised.

I think we all went into
this domain hoping to prove

that oseltamivir had benefit.

We were more likely to say that
oseltamivir had no effect, um, given

sort of the clinical practice and the
feelings of clinicians around the world.

think any of us necessarily
expected a finding of inferiority.

And so at that point, we wrote and
quickly put together a statistical

analysis plan so that we can report
these results as, as soon as possible.

We presented them publicly at the
Critical Care Reviews meeting in early

June, which were sort of preliminary
results without all patients having

achieved their primary outcome.

And as you mentioned at the beginning,
the results are now available on a

preprint server for folks to look at also.

And what we found is that oseltamivir,
um, was not effective at reducing

ninety-day mortality in critically
ill patients with, with influenza.

Um, and in fact, um, there was a
probability that it was associated with

harm or higher mortality in both the five
and ten-day arms compared to no antiviral

Scott: Okay, Lindsay, do you wanna
put some probability statements on

what Srin just referenced in terms
of harm, futility, estimated effect

Lindsay Berry: Sure.

So, so the outcome was 90-day mortality.

Um, the observed rates for the
different groups were about 19% and

20% observed mortality for the five-day
and 10 days oseltamivir groups.

And in the no antiviral control, the
observed mortality rate was 13.7%.

In our adjusted analysis, the
adjusted odds ratios for the two

oseltamivir durations were two point
one three and two point one seven.

So that's, you know, one would be no
difference in the odds of mortality.

So this is a very sh- you know, high
increase in the odds of mortality.

The Bayesian model gave 80…

Sorry, 98%

Scott: 98 and 99% probabilities of harm.

Lindsay Berry: so that the odds ratio
was higher than one, um, and high

probabilities of futility, which we
defined as an odds ratio of greater than

zero point eight three, which was our,
um, clinically meaningful odds ratio

Scott: 0.83,

which was our, uh, clinically
meaningful odds ratio Okay.

So 98% probability of increasing mortality
in this severe influenza population from

the primary analysis, uh, uh, result.

Uh, a s- a fairly stunning,
uh, sort of result given, given

the backdrop of all of this.

Okay, what, what about the result now?

Um, and, and maybe the first part of
this is, this is a fairly complicated

trial because some patients are
randomized they're eligible for

control, we talked about that.

Some sites do not have a control.

in our approximately 500 patients, some of
those are randomized between five and 10

days, but were never eligible for control.

Lindsay, you described in the
big overarching model, we have a

factor in there that is, are you
eligible for the control, which

appropriately adru- adjusts and gets
the right contrast for that effect.

But we also did analyses where you
excluded all of those oseltamivir patients

that were ineligible for control, and
are the results different for that?

Lindsay Berry: Well, I think I forgot
to mention that adjustment, but I meant

to, so I'm glad you imputed it for me.

Um, so yes, the model itself has a
parameter, sort of an indicator, was

this patient eligible for control?

So we are adjusting for the fact
that there could be differences

in patients that, um, were at
sites that offered control or not.

Um, and the model also
includes site effects.

So there's sort of two ways of
trying to address the fact that

patients, um, could have different
outcomes at, at different sites.

Um,

Scott: Um,

Lindsay Berry: so we report, uh,
the results in what we're calling

this co-eligible population.

So these are the subset of patients that
are eligible for control and also tamivir.

And the median adjusted odds ratios
are very similar, so that if anything,

they actually are slightly higher.

Um, but the probabilities of harm are
still s- um, above ninety percent.

So, you know, the qualitative takeaway
is the same in this restricted population

of patients that are eligible for control

Scott: um, above 90%.

So, you know, the qualitative takeaway is
the same in this restricted population of

patients that are eligible for control.

And there was actually a s- a,
a stunning similarity of the

mortality rate on oseltamivir at
sites that didn't have control.

It was almost exactly this 20%.

something we haven't talked about, and I
think it's, it, it's because it seems not

to be an issue, but Srin, there's really
no difference between five and 10-day

results in, in any kind of meaningful way.

Yeah

srinivas murthy: Yeah, not at all.

Um, and we looked at that in
a variety of different ways.

at it, um, with the two arms
being nested with each other,

which was our primary analysis.

if we looked at them independently or if
we looked at them as a pooled estimate.

largely, um, those aligned pretty closely,
um, telling us that the five- and ten-day

arms were not that different from each
other, which is useful information,

um, on top of the primary finding

Scott: Okay.

Yeah

Tom Hills: to reflect on the results
in the clinical context that the trial

was conducted because if you'd known
that harm might be the finding, you

really wouldn't want sites randomizing
between one harmful intervention

and another harmful intervention.

Uh, but, but that wasn't where we
were when we were designing the trial.

We were designing a trial in
the context of most sites using

oseltamivir for most patients.

And I think

Scott: I think it weighed down the benefit

Tom Hills: of oseltamivir compared
to no antiviral control, we'd be

really grateful that we had the
five versus ten-day question.

Maybe that question
would still be in play.

Maybe we'd be trying to work
that out now by continuing to

randomize between those two.

Uh, but that wasn't the context,
and some sites couldn't participate

in the no antiviral control.

But I, I don't think
we've stated yet that that

Scott: That was a minority of

Tom Hills: our

Scott: our sites.

Tom Hills: So most

Scott: sites were on

Tom Hills: with randomizing to
the no antiviral control, but a

significant minority were not.

And so it's important that the
statistical model can account for that

availability of the no antiviral control.

And it's really important to look at this
co-eligible population, uh, analysis to

Scott: Okay, so let's, let's talk
more about clinically what this means.

I, I mean, to some extent, the experiment
itself, they-- the, the sample sizes here

approximately, know, 300 patients were
eligible for control overall, uh, just

under approximately 500 patients overall
with this 98 pro- probability of harm.

What do you-- what…

You know, take us through what this means.

What, uh, Srini, what does this mean?

srinivas murthy: Yeah.

Um, and I think that's the question for
the larger community to wrestle with next.

But I think in my practice, I-- my current
hospital says every patient who's in the

ICU should be getting oseltamivir, at
least they did, and I think that needs

to change and that will likely change.

Um, so that oseltamivir is no longer
routinely used in the critically

ill population with influenza.

and this is the, I'll remind
everybody, the first randomized

trial in this population.

And so if we went back twenty-five
years and had this result then, and

how people could interpret it going
forward, and would oseltamivir be as

controversial in this population or not?

Um, probably not.

It's just because it's so established
in clinical practice that this result,

That's areas

may ruffle a few feathers and may, um,
make people ask a few questions about

some, um, of the results more broadly

Tom Hills: I think there's, um
Two questions for clinicians

and guideline groups.

One

Scott: One is,

Tom Hills: do

Scott: do you think
post-COVID there is anything--

Tom Hills: And

Scott: and the other is,
do you think it's now?

And there are some
pretty different reasons

Tom Hills: And in New

Scott: New Zealand,

Tom Hills: uh, and other southern
hemisphere sites, we have a bit

more time pressure, and so we're
grateful you presented the results

early at CCI Belfast because flu

Scott: Louis PA.

Tom Hills: In

Scott: In my hospital,

Tom Hills: our

Scott: now making guidelines to

Tom Hills: recommend oseltamivir
for critically ill patients, and

it's already shifted to say it's not
recommended for critically ill patients.

And, uh, you could make that change
if you interpret our evidence,

um, as, uh, indicating there's no

Scott: There's no DNA

Tom Hills: purely

Scott: because you think

Tom Hills: benefit.

Uh, but way to look at our

Scott: There is a, there is probable

Tom Hills: that there's probable harm.

Uh, and so some might say that
our probabilities for harm don't

meet some threshold they have in
their mind, particularly if you

look at the co-eligible population.

But I think it's pretty clear that
there's no benefit oseltamivir.

And when there was no randomized trial,
evidence before, it might be that's enough

to shift clinicians or guideline groups.

And it might be that
there's always this question

Scott: there that

Tom Hills: definitely harmful?

Our

Scott: there's always this question
about is it definite or not?

Our results

Tom Hills: uh, show a high
posterior probability of harm.

But clinicians have to interpret
that, uh, when they are familiar

with using this drug and haven't,
uh, felt that it was causing harm.

And so I think it's fascinating
to think about, you know, how

could this drug cause harm?

And it's amazing when you find something
like this in a clinical trial, you realize

there's so much you didn't know about a
drug you thought you knew a lot about.

I have to admit,

Scott: when I am,

Tom Hills: had-- I

Scott: and I've missed a couple
of days of medical school.

I didn't know that we had human
urokinase enzyme scanning,

Tom Hills: though oseltamivir

Scott: most cell can be a
good one when it's designed

Tom Hills: to inhibit the influenza virus
neuraminidase, it seems like it, it, it

Scott: It can anticipate

Tom Hills: the human neuraminidase
too, which is important in

Scott: I guess, a biology and there are,

Tom Hills: there are

Scott: you know, animal studies,

Tom Hills: are

Scott: there are

Tom Hills: randomized trials where,
where the hematology doctors have worked

Scott: doubt that if you

Tom Hills: randomize half the

Scott: patients who have an

Tom Hills: immune

Scott: problem with their platelets
and their platelets are being

Tom Hills: are being

Scott: destroyed by their immune system,

Tom Hills: If they

Scott: they randomize
patients to oseltamivir.

The people who got oseltamivir's
platelet count recovered.

Tom Hills: Uh, and

Scott: and that's because there's
an effect on human hematopoiesis.

Tom Hills: And

Scott: So is it impossible

Tom Hills: there's

Scott: that

Tom Hills: an effect here of a medicine

Scott: effect here of a medicine
we thought was safe, we thought

Tom Hills: thought

Scott: only worked

Tom Hills: the

Scott: the virus?

I

Tom Hills: because

Scott: it works on human

Tom Hills: enzymes

Scott: or

Tom Hills: some other

Scott: other pathway
that we don't understand,

Tom Hills: And

Scott: and that may explain it.

But let's be realistic here again

Tom Hills: We know that patients in
the community don't really experience

hazard, but I think it's really
important to acknowledge that patients

with flu in the community And even
patients with flu on the ward are very

Scott: doesn't

Tom Hills: to patients in
the intensive care unit.

such a

Scott: champion

Tom Hills: heterogeneity of, of clinical
illness severity that it seems plausible

the drug would work differently
across that illness severity spectrum.

And we might not understand it,
but perhaps it is plausible,

even though before this result
came out, I wasn't aware of

Scott: aware of some of the biological

Tom Hills: and pharmacology
of oseltamivir.

It's gonna be interesting to
try and unpack that, uh, over

the coming months and years

Scott: So, Srini, the, I found
the subgroup analyses fascinating

actually in the trial, which
you should all go look at.

Does that back up a little bit of
this in, in community ambulatory fresh

infected subjects, there's benefit of
this, and as you go to this extreme,

we're starting to see potential
harm, high probability of harm.

The subgroup analyses seem to back this up

srinivas murthy: Yeah.

And sort of echo a lot of what Tom
has just said, and the heterogeneity

of patients with influenza is huge.

even within the ICU population,
the heterogeneity is huge.

And so within our trial, we randomized
patients who are in shock, so they're

on vasopressor-type drugs to keep
their blood pressure up, and in shock.

We randomized patients who
had symptoms within five days

and symptoms than five days.

We randomized immunocompromised patients

Scott: Patients didn't have that.

When we randomized patients,

srinivas murthy: that.

And we

Scott: we thought that

srinivas murthy: patients who we
thought had a bacterial infection

at baseline and those that did not.

And in every one of those subgroups, the
group that was considered, quote-unquote,

"more sick," the immunocompromised,
those were proven bacterial infections,

the ones in shock, the ones with a
longer duration of symptoms, all had

a worse treatment estimate effect with
oseltamivir compared to the less sick part

of their subgroup, which is fascinating.

Um, and I love clinical trials like this
where it sort of asks and brings out

even more questions about host biology
and how, um, a virus and a very sick

individual And you're giving this drug
that you think is doing one thing, and

it may not be doing the thing you're
doing, but it does open up all these

different hypotheses about why these
patients become critically unwell and

of brings out so many questions about
what we do in critical care more broadly

Scott: And

Tom Hills: we didn't design our
trial to answer the question of

what's the mechanism of harm.

You know, we were trying to randomize
people to an intervention that

we, thought might benefit them.

And we were, we were worried there
wasn't evidence, uh, and hence

we had the no antiviral control
at sites who had equipoise.

But we weren't sort of trying to tease out
any mechanisms of hazard in our design.

So unfortunately, we don't have
a whole lot of data, to answer

the question of, you know, what
is the mechanism of hazard here?

Scott: I-- so Lindsay, this was
interesting in that because of the

heterogeneity of, of who we enrolled
here, the covariate adjustments

mattered, uh, i- as the trial results.

I qualitatively the, the, the story's
all the same, but the story became

more precise with the, the covariate
adjustments, which is a relatively

important part of the primary analysis

Lindsay Berry: Yeah, and I don't know if
we mentioned this earlier, but, uh, when

you look at table one, there are some,
um, imbalances in some of the baseline

characteristics across the arms, and some
of them sort of s- favor oseltamivir,

so maybe would be healthier patients on
oseltamivir, and some are the opposite.

Um, and so I think

Scott: I think

Lindsay Berry: it was age.

Um, the median age was slightly younger
on control, but then I think things

like invasive mechanical ventilation,
those rates were higher on oseltamivir.

And Srin or Tom, were there other
ones that, um, people mentioned

Scott: I think things like invasive
mechanical ventilation, those

rates were higher on oseltamivir.

srinivas murthy: ventilation, which

Scott: in the

srinivas murthy: higher in

Scott: oseltamivir groups, um, which,

srinivas murthy: um, which,

Scott: out as an issue.

srinivas murthy: brought

Scott: and I think the covariate
adjustment, as Miro mentioned, um,

srinivas murthy: the covariate adjustment,

Scott: really

srinivas murthy: mentioned,

Scott: lifting here.

srinivas murthy: does a

Lindsay Berry: Yeah.

So it was that, um, those imbalances.

And then I think another observation is
just if you look at the raw mortality

rates and calculate an odds ratio, you
get something that's about one point five,

whereas the model with its adjustments
results in an odds ratio of two.

So sort of where, where is
that difference coming from?

And I-- you know, while I've been
sort of meditating on the result,

I think it's, one, the imbalances
in covariates and the model

adjustments is part of the difference.

And then also just the marginal unadjusted
odds ratio is just an inherently different

quantity than the conditional odds ratio.

And, and we know even in a perfect
randomized trial with no confounders, that

marginal unadjusted odds ratio will be
closer to one, um, if there are covariates

that are highly predictive of the outcome.

So I think some of this is just
to be expected because we're

estimating an odds ratio, and then
some of it may also be the fact that

there are these chance imbalances
in covariates across the groups.

It's hard to tease out exactly how
much comes from each, each part of that

Scott: of that sort of
communication part of that.

There's an interesting part to this.

So Tom, you laid out and, and Srin
this idea of the even within the

severe, the more severe do worse.

We've got this understanding that
at some level of, of healthy, that

there's some benefit to this treatment.

REMAP-CAP is enrolling
oseltamivir before this result,

oseltamivir in moderate patients.

these are hospitalized patients that don't
meet the definitions that Srin talked

about as severe, and we're enrolling that.

So here's this of likely
mortality harm in severe.

And at the end of the day, we--
at, at the end of the left-hand

spectrum, there's some benefit of
this in, in, in healthier patients.

Now we've got this one step healthier
and a controversial or, or, you

know, decision on whether we keep
enrolling, and REMAP-CAP is enrolling

in this in moderate patients

srinivas murthy: Yeah.

Um, and I think we've polled a lot of
our sites and asked a lot of people

as to whether we've answered the
question in moderate state patients.

and like you've mentioned, at some point
along the spectrum of illness where it has

benefit in outpatients and possible harm
in critically unwell patients, there's

going to be a line where that flips.

Um, and where that line is is
something we're hoping to figure out

with continuing to recruit patients
who are moderately well unwell.

And we'll look at the data as per
our usual processes with our DSMB

very careful, um, about what we're
finding and making sure that they let

us know if it's gone in a direction
similar to the critically unwell group.

and so more to come on, on that respect
on the role of oseltamivir in other

hospitalized populations, including kids.

And I think pediatrics is an
area where I think REMAP-CAP,

um, really thought about how to
integrate the data that accumulates.

And we do interesting things where we
borrow from one population another.

And so severely ill children with
influenza are an understudied

group as most people know.

And so figuring out whether it works
in that group also is something

we'll continue to evaluate.

Um, once again, stopping rules and
DSMB close monitoring and all of those

safety issues integrated throughout.

so for everybody, more
to come in both of those

Scott: And this result is 12 and older,
I think is what we refer to as adults.

REMAP-CAP is enrolling, though our, our
enrollment rate is quite small, uh, less

than 12, uh, uh, patients, pediatric.

So, eventually we'll get a
readout of that, uh, a- as well.

Okay.

Um, uh, you know, w- w- within
this setting, Tom, w- you

know, does this have an impact?

What, what is your thought?

Now, you're reading your
colleagues a little bit.

this somewhat dismissed because
I've been using oseltamivir for

years and I never noticed it?

Of course, in a single arm anecdotal
setting, would you ever notice it?

But does this have an impact?

What, what-- Predicting your
colleagues, what do you think?

Tom Hills: Well, we're in the thick of
it with the flu season, and it feels

like the critical care community are
embracing the result and changing

practice, uh, here in, uh, in New Zealand.

And I think there are bigger
questions for the ward clinicians.

You know, does my patient look more like
a, an ICU patient or a community patient?

Because in the you know, outside of ICU
hospitalized population, there's also

a dearth of clinical trial evidence.

I think there's one RCT of oseltamivir
directly compared to no oseltamivir

in maybe seventy-six patients, and
no suggestion of hazard, but, you

know, only seventy-six patients.

and so most of the discussions
we've had have been about,

wow, what's the mechanism?

Uh, what does it mean for
those patients on the ward?

and then a little bit about the, the

Scott: that end in the trial, the,
the, the size of our REMAP-CAP trial,

Tom Hills: the

Scott: bigger group, the
co-eligibility population,

Tom Hills: which I

Scott: which I think is worth
reflecting on because we have published

Tom Hills: many

Scott: results from REMAP-CAP with,

Tom Hills: you know, a thousand,

Scott: 1,000, 2,000

Tom Hills: thousand people, uh, in a
domain, and here the number is smaller.

So some people might say

"It's

too small for you

to say

that there's an increase in mortality."

Whereas another

interpretation might be

that's a big effect in
odds ratio greater than

two.

The

trial is the right size and derived
efficiency from covariate adjustment

and things we've talked about, and
so we've stopped at the right time.

And I think that's, um, that's
an interesting sort of debate

and an interesting thing to
meditate on, as, as Lindsay said,

because, um, yeah, clinicians

are surprised by the finding of harm.

Uh, and of

course, you always wish you had more

patients give

to give more statistical certainty,

but you also don't wanna keep enrolling

to a trial

where

the accruing data suggests
you're causing harm

just to give more
statistical certainty that

this drug definitely

causes harm.

So I think there's been, uh, a shift
already in, in most intensive care

units while acknowledging surprise and
that the mechanism's poorly understood

and a big question for, for the ward

Scott: Shrimp?

srinivas murthy: Yeah, and I think
the, this idea of it being too small.

influenza studies are hard to do.

As Tom mentioned, there's very few
inpatient influenza studies, um,

because they're-- it's a seasonal
disease, it's unpredictable in

terms of its populations and so on.

And so getting an appropriately
sized study has always been difficult

across many decades of research.

And so we, randomizing almost
five hundred patients in and of

itself is, is substantial first.

And second, I think the issue of whether
we, quote-unquote, "stopped early,"

um, is a one that I'm sure, Scott, you
have many thoughts on as, uh, someone

who writes a lot about stopping early.

But remember, our goal
in this study was not to

Scott: Whether it's harmful

srinivas murthy: show whether it's

Scott: or not.

Our goal in this study, in this study
was to show whether it's effective.

srinivas murthy: was to show

Scott: And I think

srinivas murthy: and I think
that has been fairly clearly

Scott: regardless interpretation

srinivas murthy: of your
interpretations of when we stopped

Scott: I, I'm also just incredibly
struck by, you know, 500 patients being

small when there's zero randomized
evidence of this ever before.

It's all observational, and all of a
sudden, you know, that relative to zero,

that, that's big, uh, in the setting.

So it's, it, it's sort of
fascinating the reception, uh, to it.

And it…

And the other part of it is no study
ever of oseltamivir has shown a

mortality benefit in any study ever.

And I know we got a result in
steroids, for example, that was…

I, I don't even think it was
controversial, 'cause st- some studies

have shown benefit, some have shown harm.

But at least there have been some
studies that have shown benefit of

steroids, and when the result came
out as negative, it was context.

Here's the first randomized
trial in this population.

Now, it's fascinating how moderates come
out, and I know the recovery trial are

enrolling in moderate patients, so there's
more to tell of this story, and pediatrics

Tom Hills: Yeah, and

Lindsay Berry: and

Tom Hills: RECOVERY might have some
critically ill patients too, so

it'll be really interesting to see,
um, how the RECOVERY data come in.

And obviously, in the past, we've
evaluated some things that RECOVERY also

evaluated during the pandemic and sort
of very different statistical approaches.

But it's been interesting to see how the
results have aligned historically, and

so it'll be really interesting to see
what RECOVERY learn about, oseltamivir.

One thing that's fascinated me is just
how different our finding is from the

observational data and, you know, you
could run a whole series of podcasts

on, uh, reasons that randomized
trials might find things that differ

from, uh, the observational studies.

thing I've been reflecting
on is the site effects.

You know, the,

Scott: The,

Tom Hills: the

Scott: the

Tom Hills: mortality rate from influenza
at, at sites within our study or

outside of the study can vary wildly.

Uh, and so I'm always thinking
those observational studies

showed a clear association between
oseltamivir and reduced mortality.

man, the sites that use
oseltamivir routinely must be so

Scott: so different from
the science that's done

Tom Hills: from the sites that don't.

Um, and so when you're looking at, at
those site effects and then the patient

effects that might, uh, influence whether
oseltamivir is used, I'm sort of thinking

I'm really glad we've got all these
covariate aj-adjustments in our model.

Scott: So

Tom Hills: and I

Scott: I think

Tom Hills: it's just

Scott: it's just so

Tom Hills: hard with
observational data like that.

Uh, and it's, it's, it's on the one hand
hard for clinicians to shift away from

using something they've als-always used.

But as you point out, Scott, this is
the first randomized trial, so I think

people are sitting up and taking notice.

And even if they don't fully
understand the mechanism, and we

don't fully understand the mechanism,
acknowledging that this is the first

RCT data in this population and, and
responding to that, their, in, in

their bed practice at the bedside

Lindsay Berry: I just wanted to plug in
the supplement of the pre-pa- preprint.

We, um, included a sensitivity
analysis where we used

alternative prior distributions.

So

our prespecified

approach was a non-informative
prior, so we could let the REMAP

data speak for, speak for itself.

But we also considered as this, uh,
sensitivity analysis, a range of

optimistic, neutral, pessimistic priors,
um, that also varied in strength.

So maybe you come in with
a very strong belief that

oseltamivir is effective or
a very strong belief that the

effect is, is null, no effect.

So we have a, a plot there that shows how
much does this new data move the needle

depending on where you came in with,
what your beliefs were before the study.

And I think largely, you know, the results
for each prior, uh, show that the…

it's more likely oseltamivir is, um, worse
than no antiviral unless you come into

this, um, with a very strong belief that

oseltamivir was effective.

So sort of you, you started with a
99 probability of benefit, then maybe

this data wouldn't move you very much.

Um, but in that case, the sort of
prior weight is, is, is even higher

than the, the sample size of our trial

Scott: Very nice, uh,
example Bayesian analysis.

It's beautiful.

Yep, so check it out.

Okay, um, on the heels, is there anything
here or, or not worth going into as,

as, uh, in, in REM- in COVID-19, we had
multiple antivirals that all showed harm

in that virus, and I'm, I'm not aware
of a randomized trial in severe disease

where an antiviral has shown benefit.

there, is there an interesting class
effect of this result, or you don't

even really wanna go there at this
point, that antivirals, you know,

if you've seen one, you've seen one?

srinivas murthy: Yeah, it's
a fascinating question.

Um, and I think most of the antivirals
for s- respiratory infections have

been studied well in outpatients,
and very little study has happened

in severely ill, and like you've
said, we've extrapolated evidence.

COVID-19, we've shown harm
in s- with some antivirals.

The most studied one, remdesivir, where
they studied it across every severity,

outpatients, moderately unwell, critically
unwell, showing benefit, big benefit

in the outpatients, benefit in the
moderately unwell, no benefit in the

critically fascinating heterogeneity,
across different respiratory infections

with the use of direct-acting
antivirals, it's fascinating.

Um, whether I'm willing to throw every
antiviral out in critical illness

with-- due to severe respiratory
infections, uh, I'd like to see more

studies, and we're continuing to
evaluate baloxavir, for example, which

is a completely different mechanism of
action, more targeted towards the virus.

No, as we know of, host
effects whatsoever.

And so we'll have a more pure experiment
as to what the role of antiviral

effect is, um, in influenza populations
emerging, um, in the near term

Scott: Fascinating

Tom Hills: it does teach us
that we should think about the

pathophysiology in critically ill
patients as potentially different.

And I think that's kind of obvious to
clinicians at the bedside, but sometimes

we want to evaluate the same, the same
treatments and treatments we know have

worked in, in a less unwell cohort.

And, and some

Scott: Some of us are in a member leading
in domain where we're trying to evaluate,

Tom Hills: uh, interventions that

Scott: that modulate the immune response.

So shifting from pandemic virus to target

Tom Hills: response in critically ill
patients, uh, and trying to do that in

a way that's sensible and acknowledges
that there are patients with secondary

bacterial infection, immunocompromise,
and how do we think about those patients.

And I think heterogeneity of
treatment effect across the

illness severity spectrum

Scott: spectrum and various other

Tom Hills: clinical factors is

Scott: as the next frontier
for clinical trials

Tom Hills: gonna be something we're
grappling with, you know, for decades

Scott: Okay.

Tom Hills: so it's, it's a fascinating
space to be designing clinical trials and

Scott: All right.

So thank you all for joining here on, uh,
uh, in the interim for this important,

surprising, very impactful result.

Appreciate you all joining

Tom Hills: Thank you

Scott: So until next time, of course,
we'll be here, uh, on this incredible

journey of clinical trial science,
science of medicine, statistics.

Uh, we'll be here in the interim