Vasopressor Blood Pressure Targets in the ICU


>>Thank you very much, Paul. Thank you very much Rob, and
the CCR organizing committee for this wonderful opportunity
to present The 65 Trial. Definitely, very excited to be here. This trial was funded, as you know, by the
NIHR Health Technology Assessment Program, and we’re grateful for that, definitely. We’re also very grateful that the results of
the trial have been very recently accepted for publication in JAMA, and so
on behalf of the entire 65 team, we’d like to extend our thanks to Drs. Angus, Vouchner, and the JAMA editorial team. I’ll be presenting today on behalf of the
trial chief investigator, Paul Mouncey, as well as this illustrious group of individual who compose the trial management
group and trial steering committee. If I may, I’d like to start by
providing a little bit of context and explain why we felt it was
important to conduct this study. In a nutshell, the short
answer to that question is that vasoactive substances became an
integral component of routine care for shock and hypotension at a time when perhaps new
medications weren’t under the same scrutiny as they are today before becoming
available for clinical use. And I guess one way of putting it is that we
felt that the careful evaluation of the risks and benefits associated with these
potent drugs was long overdue. The other reason is that if you think about it,
when we, our daily use of vasopressors hinges on a number of assumptions that may be flawed. We tend to use these drugs
to treat distributive shock, because we worry that excessive
vasodilation cause low blood pressure and that low blood pressure causes
malperfusion, but if you think of it, these distributive shock states are
associated with high cardiac output, and it’s unclear that blood flow to
important organs is actually that low. And if it is indeed the case
that blood flow is reduced, increasing vasoconstriction may not
be the way to increase perfusion. If anything, this may actually
reduce blood flow to organs. So, we thought it was an important
area of research to tackle, but to study this well, you
need clinical trials. Relying on observational data to study this,
you end up with unremediable residual bias. And it just so happened that very
few patients had been randomized in clinical trials up until this point in time. Most of them were enrolled in the
SEPSISPAM trial that you know well, which randomized patients with septic shock
to either a higher or a lower MAP target and the corresponding higher
or lower vasopressor doses. The other trial on the left
is the ovation trial, a much smaller feasibility
pilot study conducted in Canada, which resembled in many ways the SEPSISPAM
trial except that eligibility wasn’t confined to septic shock and that the
blood pressure targets were lower. In both of these studies
interestingly, the mortality was higher in the higher math target arms, but the
difference was not statistically significant, and the interpretation of the
results went in different directions. Later on, when we pooled the individual
patient data from both studies in this individual patient
data from both studies in this individual patient data
meta-analysis, we found that the, again, the mortality in the higher MAP arm was
higher, but perhaps more interestingly, we observed that this higher mortality in the higher MAP target arm was perhaps
more pronounced in patients who were older. And so that led to the hypothesis we
derive from this observation the hypothesis that perhaps vasopressors
defined as any medication that induces vasoconstriction may be
overall nonbeneficial but in fact harmful in older patients due to their
more limited physiological reserve. And so the 65 trial asks
whether reducing exposure to vasopressors via permissive
hypotension reduces 90-day mortality in patients 65 years old or older who are receiving vasopressors
or vasodilatory hypotension. This was an investigator-initiated,
very pragmatic, unblinded trial. And lest you forget why we called it that
way, it was conducted across 65 sites, participating in the Case Mix Program in the UK. We enrolled patients who were 65 years
or older, who were receiving vasopressors for vasodilatory hypotension, as long
as the vasopressors had been started within six hours of randomization. And as long as the treating
physician acknowledged that adequate fluid resuscitation
had been completed or initiated and that vasopressor therapy was going
to continue for at least six hours. We excluded patients who were
hypotensive for other reasons as well as patients receiving vasopressors
for brain injury or spinal cord injury and patients whose death
was perceived as eminent or who had previously been
enrolled in the trial. Patients randomized to the permission
hypotension arm received vasopressors titrated to a MAP target of 60 to 65 mmHg. And those randomized to the usual care arm
received vasopressors according to usual care. The allocated interventions were
administered throughout the ICU stay, and all other therapies were at the
discretion of the treating teams. As I said, vasopressors in this trial were
defined as any drug inducing vasoconstriction, and we explicitly distinguished vasopressors
from pure inotropes such as milrinone and dobutamine, which were not
affected by the study protocol and could be used freely in both arms. Randomization was available
24/7 and was stratified by site using permuted blocks
of variable lengths. We, in this context of resuscitation for
shock, we used deferred consent model, and patients lacking the capacity to consent
at randomization were approached later on by the research teams
when deemed appropriate. Posters and leaflets were
available and data were retained up to the point of withdrawal of consent. The primary outcome was 90-day mortality. We also evaluated ICU and hospital mortality,
length of stay in the ICU and hospital, during of respiratory and renal support,
number of days alive and free of respiratory and renal support up to day 28 as well as
cognitive decline measured by the IQ code score at 90 days and at one year, and health-related
quality of life also at 90 days and at one year. A sample size of 2600 patients
provides 90 percent power to detect a six percent absolute risk
reduction from 35 percent in the usual care arm, as estimated from the Case Mix program data, to
29 percent in the permissive hypotension arm, assuming a 2.5 percent lost to followup. The data monitoring and ethics committee
composed of Professors John Norrie, Andreas Laupacis, and Danny McAuley reviewed
the single interim analysis, which included data from 500 patients and recommended
that we continue without modification. The study protocol, the analysis
plan were both fully published, and the trial was duly registered. Now, the results. The 65 teams screened for
enrollment 10,755 patients, and of those, 2930 were deemed eligible. And of those, 2600 were randomized. So, this amounts to 89 percent of all eligible
patients actually randomized in this trial. Of the 2600 patients, two were randomized in
duplicate, 137 withdrew or refused consent, and only one was lost to followup. So, at the end of this, 2463 patients
were included in the primary analysis or 95 percent of all patients randomized. At baseline, the two groups were well balanced
in terms of age, sex, chronic comorbidities. The one exception perhaps being the proportion
of patients able to live without any assistance for daily activities, there were 65.6
percent in the permission hypotension arm and 69.1 percent in the usual care arm. Other than that, the illness severity scores
at baseline, the proportion of patients with sepsis, septic shock, or
without sepsis were the same. The MAP at randomization was also
very, very close as was the intensity of vasopressor therapy and the duration of
vasopressor therapy before randomization. So, but if the two groups were quite similar
at baseline, immediately after randomization, they diverged significantly in
terms of exposure to vasopressors. Vasopressors were discontinued earlier
in the permissive hypotension arm on average 10 hours earlier
than in the usual care arm. The trial did not impact the
selection of vasopressor agent, but it did change the total dose of vasopressors
administered to patients and calculated in terms of norepinephrine equivalents. The total dose was 31.5 milligrams
in the permissive hypotension arm and 44.3 milligrams in the usual care arm. And not surprisingly, the average MAPs
while receiving vasopressors was also lower in the permissive hypotension arm. Remember, those averages
were calculated strictly from hourly MAP values collected while
patients were receiving vasopressors, and the vasopressors were stopped
earlier in the intervention arm. The co-interventions were used
rather similarly in both arms. There might have been slightly more
corticosteroids used in the usual care arm. Other than that, pure inotropes and
fluids, interestingly, were seen, you know, to be used equally in both arms. The fluid balance as well as the urine output
was not different between the two arms, and this was calculated during the
first episode of vasopressor therapy. Now, the primary outcome of 90-day mortality. We found that in the permissive hypotension arm,
41 percent of the patients had died by day 90, and in the usual care arm, 43.8 percent
for a difference of 2.85 percent at 95 percent confidence interval ranging
from minus 6.75 to plus 1.05 percent and a corresponding p value of 0.15. The prespecified adjusted analysis for the
independent variables appearing at the bottom of the slide yielded an odds ratio of 0.82 and a
95 percent confidence interval ranging from 0.68 to 0.98, and you can appreciate
that in a post-doc analysis, when we measured relative risk, the
direction and general trend was the same but the 95 percent confidence
interval crossed the no effect line. On screen, you can see the survival curves. They diverged fairly early on,
about one month after randomization and never crossed thereafter,
but the difference in time to death was not statistically significant,
as you can see examining the hazard ratios at the top right corner of the screen. We ascertained whether the treatment effect
varied across different subgroups defined by seven subgroup defining variables, and the
test for interaction between treatment effect and chronic heart failure, atherosclerotic
disease, predicted risk of death at baseline, sepsis categories, and the intensity of vasopressor therapy was
not statistically significant. The interaction between the intervention and
age, again, was not statistically significant, but we found that the reduction
in mortality associated with permissive hypotension was more pronounced
in patients who had been identified at baseline as being chronically hypertensive. This analysis was not corrected
from multiple [inaudible], and should be interpreted with caution. These are the secondary endpoints. We found that permissive hypotension
did not impact the duration of advanced respiratory support, and it also
did not change the number of days alive and free of respiratory support by 28 days. Similarly, it did not change the duration
of renal replacement therapy or the number of days alive and free of renal
replacement therapy by 28 days. Cognitive decline at 90 days and at
one year assessed by the IQ Code score, which is a score that ranges from 1, signifying
much-improved cognitive function to 5, signifying much worse cognitive function
was also not different between the two arms. A value of 3 means no change. And health-related quality of life also was
not different at 90 days or at one year. We paid, you know, special attention
to adverse events in this trial. There were a total of 79
serious adverse events reported in the permissive hypotension
arm and 75 in the usual care arm. You can see those SAEs broken down in different
types of SAE that we had either prespecified on the left side of the screen or that were
reported throughout the course of the trial. So, in summary, we found that
permissive hypotension was associated with a significant reduction in exposure
to vasopressors but was not associated with a statistically significant reduction
in mortality and was not associated with a difference in secondary
outcomes or adverse events. I’d like to take the last few
seconds of this presentation to wholeheartedly thank the wonderful clinical
and research teams at each of the 65 sites as well as the patients and the families
for accepting to participate to this trial. And perhaps on the more personal note,
I feel I’m a very privileged individual as a Canadian critical care trials group member. I have the opportunity to work with some of
the most brilliant researchers in this area, but for this specific trial, the full might
of the NHS/NIHR research infrastructure and the expertise of the wonderful ICNARC
CTU was needed, both of which we’d be happy to import along with Harry
and Megan, by the way. But on a very personal note, Paul, Cassie,
and Alvin, I’ll be forever grateful for the exchange of ideas and the friendship. Thank you. [applause]>>Fantastic. Thank you, Francois. [ Applause ] So, there we have it. So, time for discussion. Anders Perner from Copenhagen
Editorial please, thank you.>>Thanks Paul. Lights please. Good morning. Thanks [inaudible] for the invite and
to host this 65-trial presentation. I think it’s an amazing day for all of us. So, I was kindly invited to give this editorial. I’ll mainly touch upon the interpretation
of the trial, which I believe that I and we in general can do more freely than the
investigator’s presentation that you just saw. All of you. This is very jumpy. Ah. In between there, there was some
kind of a joke with being age 65. Just to remember me, so huge congratulate the
investigators, the ICNARC group, Cathy, Paul, Francois, David, all 65 investigators, and
I think the full UK critical care community for producing this very important result
for us in a meaningful way, I think, and I think we will, as I’ll show, learn a lot
and potentially improved the care of patients. So, you saw this as the sort of
overall conclusion of the trial. Conservative as the investigators should be,
interpret it sort of in the present paradigm where we sought to use this
term, statistical significance to sort of dichotomize trial results. That’s the part that I will challenge. I hope I can get this to work. You did it much better than me. Yeah. But it turns double. Now it doesn’t turn at all. Ah, so this has been discussed when
the primary outcome sort of fails, which is a bit harsh word, I think, a serious
paper in the New England Journal of Medicine, they’re specifically asking the question, what
to do next when you have a so-called failure. And there is a table in that
paper listing questions to ask. Some of them apply for the
65 trial, some does not, and I think there are some important
ones that are not on this list. And I’ll touch upon those. I think this is a very important
paper that came out last year. Three statistical colleagues of
ours, 800 co-signers on this paper, named scientists rise again, and this
is the term, statistical significance. The main point of this paper is that we
should stop dichotomizing trial results. It’s too simplistic. These are the results you just saw, and
this is the obviously the main result, three percent mortality difference
between the two groups. I would say personally that I think
this is an important difference. Other have thought so as well. The [inaudible] and some of their early trials
was safe study, define the sample size based on this minimally important
mortality difference. So, I think myself and many others would
consider this difference important. So, the challenge is this. So, the p value was 0.15. Again by the current definition, this
result was not statistically significant. But what does it mean? I’m not a statistician but I’m
quite sure it does not mean that we have proven the null hypothesis. Clearly, the confidence interval gives
more information and is more and more used to report [inaudible] results
in general, also here. So, the confidence interval goes from
a quite large difference minus 7, favoring permissive hypotension and
potentially one percent mortality increase with the use of the intervention. Again, in the sort of current paradigm,
we’ll dichotomize this and say, well, the null is included in this range, and therefore the result is
not statistically significant. So, these are the point estimates that are
sort of compatible with the trial results. And by dichotomizing at null, we will make these
numbers equal, but a plus 1 is equally as likely as minus 3, for instance, but they are not. Minus 3 in this dataset is the most likely
result, and the further you get away out into the 95 percent confidence
interval, the less likely the results. So, 0 and plus 1 is less likely than minus 3. But we and the current paradigm
with dichotomizing, we would create, we create these numbers as equal. They are not. So, I think this leaves us in a position
where we should change where to go, both with the results of this trial and also in general aromatase inhibitors
how we interpret to our results. There are, obviously, many
more issues to consider than the simple statistics of the trial. Did trial result come out at the trial
with internal and high external validity? I think yes for the 65-trial. That’s important. Was the primary estimate, the minus
3, supported by sensitivity analysis, subgroup analysis, secondary outcomes reported. There were, so there were a number of missing
valuables for the primary outcome as prudent. Worst case, best case. So, now your analyses were done, and
they supported the primary result. There was a adjusted analysis
for potential baseline imbalance, again, it supported the primary result. You already saw this, the
number of subgroup analysis, all point estimates here favor permissive
hypotension except maybe patients with no sepsis, maybe. But the vast majority favors the intervention. So, again, supporting the primary result. Importantly, as you saw, one of the secondary
outcome measures were time to death of one year. The lines do not cross, so this result,
again, supports the primary result. So, we can mark those important
considerations as well to the 65-trial. Is it plausible, is it plausible that the use of the intervention reduces
mortality by three percent absolute? I think it is. The trial was powered to show a six
percent reduction, so this is half. Another mark for the 65-trial. I think there is a case for interventions
where it means that we do less. I think if you want to do more, there is
a higher burden of proof on the shoulders for those who want to do more, in this
case, reducing the exposure to vasopressor. We are doing less, so I think
the 65-trial marks this as well. We will do less if we incorporate the
intervention into clinical practice. Finally, no trial result can stand on its own. Therefore, it should be incorporated into
updated meta-analyses, preferably of trials with low risk of bias, and
obviously, this has not been done yet. The results are presented
here for the first time. I’m quite sure that the, oh, I would be very
surprised if the investigators are not working on an IPDMA, so an update meta-analysis. And then we can update guidelines, and all
these processes are extremely important for obviously clinical decision making. Oh dear. [ Inaudible Comment ] What’s that? [ Inaudible Comment ] Waiting. [ Inaudible Comment ] Loading. All right. So, to sum this up, [laughter] so I think this
trial posed a highly relevant clinical question. I think the hypothesis behind was
supported by good clinical data. I think the investigators
are to be congratulated to perform a large, lower risk-of-bias trial. I think the results, as I show, is very useful. They come with some uncertainty. This is clinical research for you. It will likely change my practice. As said, I’ll probably wait for the updated
systemic [inaudible] to make the final decision and maybe even updated clinical
guidelines, at least our local ones. But until then, a huge congratulation. I’m sure that this trial result will
live for a long time both in itself, in updated systematic reviews,
in guidelines, and again, I think we should congratulate the
investigators, the full community, the UK Critical Care Research Forum for
this work, it points to a bright future for critical care research in
general and in the UK in particular. Thank you. [ Applause ]>>Anders, thank you very much for
that very insightful editorial. I mean, with time, are there any points that
the investigators would like to take [inaudible] at the moment, or are you happy
if we move onto questions? Any particular points you’d
like to raise at this point?>>I’m quite happy by the interpretation
of the results we just heard.>>Yeah, okay. So, we’ll move onto questions. I’d like to focus initially for the next sort of
10 or 15 minutes on the outcome of this trial. We’re going to move onto the
implications of this trial in terms of performing research in critical care. Paul is going to talk to that point later on. So, if there are any questions
on that, we might park them. Just a quick [inaudible] poll, while on call
this weekend, we contributed to this trial, and I’m a bit more impatient than Anders,
and I might not wait for the meta-analysis. So, who’s keen on introducing
permissive hypotension right now in patients of 65 and over? Could you vote please? Because my trainees are going
to find out [inaudible]. Okay, who wouldn’t and wait for a meta-analysis? Okay, thank you very much. That’s very helpful to my
practice this weekend, thank you. [laughter] So, congratulations. Practice changing. I guess on of the generalized ability questions
that I was considering about applying this in my own practice was obviously there
are different causes of hypotension within the intensive care unit that I
might be looking at sedative practice, the use of regional blockades, or indeed the
causes of hypotension in the first place. Now, you’ve done a big trial here, and I’m guessing you’re relying a
little bit on the randomization process. But did you record any data in a trial that you
were trying to record as little data as possible that might help me with that question?>>We were, we carefully ascertained whether
patients– should I, do you want me to–>>Please, yeah.>>We carefully ascertained if the patients met
sepsis criteria as shown and did not find that, you know, that significantly
modified the treatment effect. And we also meant to include patients who
the physicians believed to be, you know, experiencing vasodilatory hypotension
and not specifically sepsis, because the underlying hypothesis wasn’t
getting at a specific effect of a specific drug for a specific illness but rather the risk
profile of a class of medication in a population that had been enriched as perhaps
being more prone to adverse effects. So, other than identifying sepsis
and categorizing sepsis into sepsis and septic shock, we haven’t got the data
to date going to more specifics, I guess, but perhaps in the future with [inaudible].>>Thank you. Yes?>>I have to say that this was
an efficiently designed trial where we tried to mesh within the [inaudible].>>Yes.>>So, we’ve done a full linkage of all
the patient’s within the case [inaudible], which is the national [inaudible]
the physical care units in the UK. So, if there’s some post-doc insights, we
will certainly be trying to make the full use of the data we both collected and we’ve got
available through routine linkages [inaudible].>>Yeah.>>Do you want to stand up here maybe? [ Inaudible Comment ] Yeah, but [inaudible]. Rob likes us to be nice and relaxed, so we’ll
perhaps all sit down, and I should do so too. So, in terms of the audience, I mean lots of
practicing colleagues out there, trialists, and people that were engaged
absolutely in this trial. Are there any questions from the audience? So, we have a question here, Peter. Do we have the microphone? So, we have one question here, and yeah.>>Peter McLawson, Plymouth. Thank you for the very interesting trial. My question is, did we really
achieve permissive hypotension? The mean pressure, I think,
in the intervention was 67, which surely that’s what are
the current guidances to aim for a [inaudible] pressure of 65. I think the challenging thing is when you set a
target blood pressure is to actually achieve it.>>Can we have this, I had the few extra
slides on that first presentation [inaudible].>>Just to bring that in, what I would say,
sir, really what the aim of the study was to try and reduce the exposure of vasopressors
rather than actually necessarily going, we think a low blood pressure is the key to
this is actually about reducing that exposure. And as you saw from the baseline tables,
the blood pressure, the map that’s coming into the study, when the clinician expected
the patient to have at least another six hours of vasopressors, it was around 70. So, actually on the permissive hypotension
arm, if you think a perfect adherence to that arm would be coming
straight off vasopressors, and that patient would have had a mean
arterial pressure of 70 for the trial. And actually, I think the idea that we
tried to hit 66, it was a guide to try to reduce the exposure to vasopressors. So, if everybody came in
[inaudible] everybody came off, you would have seen a mean arterial pressure, a mean mean arterial pressure
of 70 throughout the study. But actually, it wouldn’t necessarily
have meant that we didn’t meet our aim, which was trying to reduce
that exposure to vasopressors. There is, first of all, there is some quite
heavy evidence of how the practice did differ from the first hour, I’ve actually,
there just wasn’t time to present it, the first hour of patients coming onto
the study, it was quite copyright law in the permissive hypotension arm the dose was
reduced early, and the blood pressure came down, and in the usual care arm, actually the
dose kept going up in the vasopressors, and the blood pressure was going up very
early, and you could see the separation. And there is a couple of [inaudible].>>And so, yeah, maybe there is
no need to, but the bottom line is because those MAP averages were calculated
only for hours receiving vasopressors, if the vasopressors, if the intervention was
followed, then vasopressors were discontinued, and the MAP was, you know, gave a
false impression of being elevated, where in fact the intervention
had the desired effect. What the data that Paul is alluding to is this. These bar charts show you, on the y
axis, the hours of vasopressor therapy, on the left in absolute terms,
on the right in relative terms, and you could see vasopressors were
administered longer for longer periods. And the top part of each bar are
hours where the map was above 75, and vasopressor dose rates were either
stable or the dose even increased. And you can see that there were more occasions
of these vasopressor infusions not being reduced as stipulated by the protocol in the
usual care arm than in the intervention. And then as you go down in the bar charts are
those modifications of vasopressor infusions that were appropriate according to protocol. And so, on balance, the intervention
increased the attention paid to reducing or discontinuing vasopressors, and another
way of looking at this is by looking at the hourly modifications, the vasopressor
dose rates, and the corresponding change in MAP for the first 24 hours on trial. So, very, very quickly those numbers
referred to hours after randomization. The two groups diverge pretty
quickly and pretty significantly.>>Cathy?>>Can I just, so one of the things that
convinced me not only obviously the IPDMA that Francois had led [inaudible] of SEPSISPAM
was actually looking at some UK and HS data that we had from some of our
previous trials, daily data. And really, what the purpose of
the trial was for me was about sort of what I might call curbing drift, if that
makes sense, and what we had was daily data that showed that patients are
being put on vasopressors early on in their sort of critical care stay. And then, we just happened to have some
trial data that showed daily what their kind of mean arterial pressure was and the
fact that they were still on vasopressors. And the values were going from
sort of 65 to 70 to 75 to 80 to 85, and they were still on vasopressors. And so, it felt to me that the use of permissive
hypotension here wasn’t an absolutely, but it was about trying to curb behavior. It was like a behavioral change intervention. What we really wanted to do was try and
reduce the exposure to vasopressors. That was the aim of the trial. The permissive hypotension target was one of
essentially trying to change that behavior by sort of curbing this drift, almost sort of
getting patients off or not increasing doses, sort of, but that appeared to be sort of what usual practice looked
like from the data that we had.>>And of course, we saw that was very important in the adrenal trial, getting
off vasoconstrictors. Paul, I think you had a question? [ Inaudible Comment ]>>Here we go now, you can hear me. What I’m wondering now is it’s all well and good
to say we achieved separation of blood pressure, we achieved separation in vasopressive doses,
but, you know, I’ve always been writing on the chart met greater
than or equal to 65, right. So, I still don’t know if I write
met greater than or equal to 60 that that’s actually safe if people do it. And so, I wonder what the thoughts were of
the investigators on what the default target that gets prescribed should be and whether that should actually change
in response to these data. You know, like it’s easy to say, well, you
know, if I go around and I see the patient and the blood pressure is a little bit lower, I
might be a bit more relaxed than I was before. But actually, the target,
what’s your take on that?>>Well, so the targets that were prescribed
in the context of this trial were MAPs, you know, with a range between 60 and 65. That’s how it was prescribed, and then you saw
the blood pressures and how it was delivered. But how it was written in chart is 60 to 65. Now, I’ll assume that when you write,
what you write in the medical charts, then you also obtain something
that might be slightly different. In terms of how to write the prescription,
that’s how it was written in the context of the trial, and for about 50 percent of all
hours on vasopressors, the MAP values were between 60 and 65 in the intervention arm. That provides some reassurance. But is there work to do to
connect what’s written in the chart with what’s actually delivered to the patient? Definitely. There was a lot of– but I’d say that that
work had been needed for quite a while.>>And so my followup would be to say
well surely to an extent one of the things that stops the blood pressure actually
going down to 60 in a substantial proportion of the patients, there’s a feeling from
the commotion at the bedside that actually for this particular patient in
front of me right now, I actually, I don’t think a MAP of 60 is okay. So, I’m going to make the blood pressure
just a little bit higher than that. It might not even be the doctor,
it might be the nurse, right, who is titrating the pressors up and down. So, the difficulty, I think, that we might
face here now is that disquiet might be removed by the results of this trial, and then if people
try and implement it, then all of a sudden in clinical practice the MAP might end up being
60, and yet we don’t know if it’s safe or not.>>I think what we found was that there was
very low duration when aiming for 60 to 65. What people tended to do was aim
towards the top end of that range, like clearly to give themselves
a little bit of breathing space. And we saw very few measurements below 60. I think it was just over three hours sort
of per patient that were below 60 overall.>>Yeah. And so point well
taken, and that may be the case. I think that on balance and
historically, and you’re right, this is very much a nurse-driven
intervention, and it will continue to be so. So, whatever you might want to
do to your practice, you know, reading this, you better talk with the nurses. But the point, I think, is that
collectively clinicians, physicians and nurses are extremely concerned by the
risk of hypotension, and I doubt that all of sudden MAPs will fall to very low numbers. I think that we have not been very
concerned by the risks posed by, you know, significant exposure to these potent drugs, and
if anything, like there’s never an upper alarm. You said, you prescribed this
as a greater or equal than, and that’s interpreted as a threshold. You know, never go below but as high
as you want above and that’s safe, and we like it when the alarms don’t go off. So, we’re very much high up there to
make sure that the alarms don’t go off. So, if anything, this gives us reason to
pause and perhaps examine the possibility that it’s not that safe way up there, and I
doubt that we’ll be in a territory where all of a sudden we’ll have crazy hypotension.>>So, time is moving on, and
we do have to progress this. I don’t want to curtail this really important
discussion, but we do, two quick points perhaps from Anders and then Morton, and
then I want to invite Paul on stage. So, just briefly.>>Yeah. So, I think it’s a very valid point,
and so this is obviously a complex intervention. And it’s setting a new target, so 60 to 65,
aiming at producing exposure to vasopressors. So, that’s the intervention. It was done in 65 UK ICUs
with the culture that’s there. If you want to transfer the results to
your own ICU, you should mirror that. Is my culture and my ICU the
same as these 65 ICUs in the UK? If you can see that, I think the
results are directly transferable. If not, there may be a problem, and you’ll have
to discuss that locally what to do about it. So, I think the point is very valid.>>Thanks Anders. And just finally, Morton,
did you have anything to–>>Morton, Copenhagen, Denmark. Thank you for a very nice
trial, and good presentation, and I also think that it has
clinical implications for sure. So, do you think, are the findings
unique to patients beyond 65 years? Could this apply to patients age 50 too? Is it a general finding?>>I think we need to find out. Yeah.>>We really used age to try and enrich
the population to deliver a study that we thought there was a population
that would most likely benefit from this. And actually, it’s quite interesting looking
at the [inaudible] of the interaction with age [inaudible] go higher
in younger adults. We don’t know enough.>>So, we’d like a trial that’s
hypothesis generating [inaudible].>>Yep.>>So, fantastic. So, one of the important things both to the
UK and to the international community is about how this trial was
conducted, as we’re about to hear, and may produce further discussion later on. And I’d like to invite Paul now, Paul Mouncey,
who is the director of resource at ICNARC, to come and talk a little
bit about trial delivery. Thanks Paul.>>Thank you, Paul. Oh [inaudible]. And thank you to Rob and
the organizing committee for letting us help design this session, and
this is something we really wanted to sort of highlight and really how the UK
community delivered this trial is something that Francois very kindly
avoided in his presentation. Okay. Okay. So, we’ve got the title of How We Did This. So, there are a number of sort of people who
took part of 65, but really where we want to focus is on the sites and
really the delivery at the sites. And I think we get to stand up here and
talk to you and tell you the results, but actually in reality, it’s the 65
hospitals who delivered this trial who need the congratulations rather than us. Okay. This is skipping along
quicker than I expected. Okay. So, recruitment, as we
can see, is difficult in trials. I’ll just see if I– I’m sorry. The animations don’t work. So, it’s critical care, the very complex trials. We can see on these two trials,
the calories trial is the top left, looking at what was the original
target recruitment. Promise is the bottom right. So, calories is looking at routes of nutrition, which have a 36-hour time
window for recruitment. Promise, looking at early goal-directed therapy
[inaudible] a similar six-hour time limit for recruitment, 65 did. The black lines reach to the top of the
target, and the actual, which are the blue and the yellow lines in the respective studies. So, as you can see, Promise is
around, I don’t like to say it, but about 18 months behind sort
of where we hoped we would be. But then we got to, I’m hoping
the animations do work here. Ugh, the 65 trial where if you could see it,
the black line was the pretrial estimate. If we went out to 2600 patients
from what we initially anticipated, we’d be about to finish recruitment next month. So, it great to be able to get up and present
the results before we actually would have expected to finish recruitment. You can see, the UK community
delivered the actual recruitment well above what we expected going into the study. And it was based on quite robust
data on the cases program audit. So, we knew how many patients were out there. So why, I’m having Anders’ problem. I’m not sure why that is. Oh, okay. So, why. So we looked, so looking again, it’s between 65
on Promise and still a little bit by comparison. They’re both quite similar. They had six-hour timeframes to be able to get onto recruit the patients
from admission into the study. We see it’s looking at day
of the week of randomization. So, Monday to Sunday. You see 65 is pretty flat, slightly lower
on a weekend, but you can understand that. Promise, very few patients
actually recruit on the weekend. Very much, this is all around trying
to embed the trial recruitment and procedures within [inaudible] care. When you look, oh, I’m not even touching it, when you look at time of
day, it’s even more stark. The Promise trial, very few
patients recruit in the nighttime. On the 65 trial, pretty much you can’t tell
what is the night and what is the daytime. So, working from the left, it’s
midnight, so it’s midnight to midnight. X axis, and it’s pretty flat across. It really shows that patients who have
been recruited 24/7 from the sites by– it didn’t need the research
nurses necessarily to be there at the bedside to recruit the patients. The whole clinical staff were
involved and engaged in the study and recruiting the patient throughout
the night-time and the daytime when they had other things to do. And so it normally isn’t at
the top of their priority list. So, this is fantastic to see. So, that’s fine, so looking
at how this was achieved. So, the things I want to pick up
are really around the infrastructure that helped deliver the study and deliver all
critical care studies within UK, the UK and HS, and also how we helped or how the study helped
embedding the procedures into routine care. Okay. Funding. So, we might just skip the scoping
slide between the next slide. Oh, no. Okay. So, funding. We got funding through [inaudible] called
NIHR HTA, which was really around also, not just efficiently delivering a trial but
an efficient process, contract and process, etc. We had funding confirmed
only three years ago. We got the funding for the 65 trial. We had the contract signed a month later. The grant start date was the start of March, and we projected to start
recruitment in September 2017. So, it was a very efficient start to the– Okay. I’m looking at UK governance/approvals. We’ve got a streamlined UK, apart from what I
had told Anders over breakfast this morning, and we’ve got a streamlined UK approval process. We’ve got the start of the
grant was the 1st of March. We submitted for our Health Research
Authority Approvals, sort of seven days later, so we’re ready to start, ready to get going. We had ethics approval within six weeks. And also, the health research authority
approval, which coordinates the whole sort of approval process on the same day. So, we probably started the grant within
sort of six weeks, six or seven weeks, we were sort of ready to get going. Moving onto the important topic of sites. Okay. So, we had a huge interest across the
whole of the UK sites with an expression of interest form 124 hospitals to take part. We needed 65, we needed 65
for the symmetry of the study. We had a 65 blood pressure,
65 sites, and we really needed to make sure that we kept that symmetry. So, we had 101 site feasibility questionnaires. This really allowed us to pick the best sites
that we felt would really deliver the study. So, we looked past sort of quality of delivery
on our studies and other studies across the UK, sort of critical care research community. But we also looked at how enthusiastic people
were, how well they filled out the forms, how enthusiastic the team, were the whole
team built in, bought into the study, or was there [inaudible]
who weren’t 100 percent sure if permissive hypotension
was the right way to go. So, we took it as an opportunity because we had
65 sites, it was a straightforward study to try and also work with new sites that
hadn’t delivered a lot of research within the UK, within critical care before. We thought that was a really
important sort of part of the study, and we had approximately half were
non-university affiliated hospitals. So, we selected, initiated, and activated
[inaudible] Francois [inaudible] me went to all the sites to train them all within a
few months, which was a lot of travelling, especially for Alvin and Francois. And this is what we got, 65 sites across
England, Northern Ireland, and Wales. We’re really [inaudible] with
the National Clinical Order, and that’s where the Case Mix program, the areas of the Case Mix program
[inaudible] Scotland weren’t involved. And the site set-up. So, the gray bars is what we expected. We actually got set-up started
two months earlier. And the sites were very keen to get
going, and we had 17 sites open in July, when we were looking to start
recruitment in September 2017. It was a fantastic sort of buildup,
site ready, keen to get cracking. And we hit our 65 sites pretty soon
or around the 65 sites pretty soon. And the top ten recruiters, actually
I had a quick look at this yesterday, whether there were, the feeling was the smaller
[inaudible] often are the better recruiters. It’s not all about teaching hospitals,
delivering our research in the UK. And if you just look at the top ten recruiters
there, they’re not necessarily the biggest, most familiar names that you’ll see, but
they delivered brilliantly on the study. These were the sites who
really did embed the study. They got the whole teams bought into it
and [inaudible] the study throughout 24/7. I’m just picking up, we got Queen Elizabeth
Woolwich there who is at the top of the list. They’re university affiliated,
but it was actually their first– it goes back one, which it’s been doing, oh. Oh, not that far. Thank you. And they actually got a highlight by
the NIHR on their recruitment on 65. This was the first study they delivered,
and really was about how they managed to get their whole team involved
in delivery, and it’s just, it was a fantastic story to sort of see. And this was only possible because of the, these are slides from [inaudible]
the national specialty group for the Clinical Research Network. It was a fantastic job. The National Clinical Research Network
is a key part of delivery at sites. So, when we put in a grant
application to the NHR, we tell them how much it
will cost to run [inaudible]. We also tell them how much we
need to give hospitals, but also, we tell them how much the hospitals need to
actually have for all these aspects to identify, recruit, consent patients, and that comes
through local funding through the networks. And it’s just a, it’s a fantastic setup that
we know research nurses are already in place. They’re delivering portfolio
studies, and we can just tap into the research nurses to deliver that. We don’t need to set up with completely
new research nurse, etc., for new studies. And they’re brilliantly trained,
and they will deliver the studies. They know the area fantastically well. When I started 10 years ago, it was a
[inaudible] reasonably earlyish in that journey, and the research nurses were still here and
delivering and now are really experienced. And that was just to show there are
clinical research nurses across 31 sort of clinical specialties, and I say
Paul leads the critical care aspect and is a fantastic resource. A bit of self-congratulation
from the critical– oh, sorry. The clinical trials unit. We think it’s important that
you’ve got an experienced, specialized unit in critical care research. We had a really focused trial management group. We had a couple of clinicians, we had a
research nurse who wrote all the transcripts, [inaudible] wrote all the
transcripts for the research nurses for the deferred consenting model. We had two patients on our trial management
group who put in fantastic support, and it was really, really important. Alvin did a wonderful job as trial manager,
which is a huge role in these sort of studies. Francois was with us for a year with the international collaboration,
and we’re very open to it. And we got a large network of hospitals through
our Case Mix program and we could data link and routinely link to the [inaudible]
which is a fantastic resource to have. The 65 promotion messaging
was key, and using the, these are just aspects, the
65 logo was fantastic. Cathy also had the credit for designing
it, using a speed sign for [inaudible] stop at 65 was a very useful and very good thing. And people could make, these are just examples
of what the research nurses were trying to do to try and imbed and to just
raise the profile at 65. It worked fantastically well. The randomization and intervention delivery
was streamlined and straightforward. We had a simple telephone randomization system. We gave the option of online,
but it was never used. I think we had two or three
patients we treated online. Minimal training was needed. And intervention, it was [inaudible]
single parameter that could be delivered by clinical staff, and we used the 65
prompts to remind about the intervention. The relevant consent model. So, you’re not rushing people into consent. So, the importance of actually informed
consent deferred is vitally important. [Inaudible] slightly higher rate of nonconsent,
but it actually mimics [inaudible] work in pediatrics, intensive care that we’ve
done, the team in Liverpool have done, and it mimics very well that this is
really acceptable way, not causing stress, to stress the family members
and patients at the time. And the last thing I want to say is
really around the data-enabled aspect. It really takes the pressure off research
nurses to fill in too many [inaudible] forms, too much data at the bedside, and link
into routine data with civil registrations but also Case Mix program was absolutely
key, and we focused data collection on blood pressures and vasopressors. And just again say thank you. Thanks all.>>Thanks Paul. That’s great. [applause]>>A bit long.>>So, I’m going to stay seated at this point
because we now want to invite the panel. Before we do that, I just wanted to ask
Matt Devine if there’s anything pertinent that has come up on Twitter through the hashtag,
the big international audience out there, that you might want to raise at this point.>>So, Twitter has exploded this morning, and a few things have emerged,
which I’ll just summarize. Does this have unintended consequences? Will this make us less like
perhaps to admit patients to ICU who have borderline mean arterial
pressures at the point of referral?>>So, we’ll perhaps bring
that up in the discussion. If you could just pose that question
and think about that, hold that. And the second?>>And the second thing is really
around methodology and statistics, and given in this study there was
that statistical significance, but perhaps a signal to a change in mortality. Do we need to pay maybe less emphasis on p
values and look more at clinical significance?>>Okay. So, I think those might
be two good starting points to introduce the panel, actually. So, we’ve got a list of people down here
who are real experts across the world, and I just wondered if we could go around and
spend sort of 15 seconds saying who you are and what you think you bring to this. We’ve invited you here for a very good reason. Terry, do you want to start with you?>>Certainly. Hi, I’m Terry [inaudible] from the University
of Washington in Seattle in the United States. It’s certainly a pleasure to be here, thank you.>>Thank you, Terry, and you’re very welcome. Cathy [inaudible]–>>I’m Cathy Rowen [phonetic]. I’m director of the [inaudible]
clinical trials unit. I’ve got absolutely no idea why I’m here. [laughter]>>Accountability, Cathy.>>Yeah. [phonetic]>>So, Anders, we, please–>>Anders [inaudible] Copenhagen
intensivist and trialist there.>>Thank you.>>Andrew at the end.>>My name is Andrew [inaudible]. I’m a clinical trial statistician from the
University of Pittsburgh in the United States.>>Great. And we’ve already
met Paul and Francois. So, should we take the clinical question first? Perhaps Francois and Terry and Anders, around
this whole question of bars to admission to ICU, that perhaps this notion of a
patient that is a bit borderline in terms of blood pressure control. There might be an increased barrier for
admission as a result of this trial, which, you know, may prevent other
important interventions, observations, and gestalt that we offer in critical care. Francois.>>I mean, who knows, right. Are patients mostly admitted to ICU
because we want to administer vasopressors, and let’s say we are worried that
they might be in shock and choose not to administer vasopressors,
would we feel they need ICU less? I’m not sure. It’s a great question. It’s something maybe to study in
the aftermath of this publication. Then I, you know, there are a number of
other interventions, and trying to get us to recognize, you know, early warning
scores, and that’s still happening. So, I’m not sure. And then, yeah, I guess I
would say I’m not sure. It was an interesting question to ask.>>Okay. Well, thanks for addressing it. Terry, have you any thoughts from a
West Coast of America perspective?>>Yeah, no, I think that’s a very
important concept to bring up. I think when we’re generalizing
our trial results, we need to remind ourselves how was this
done within the setting of the trial, and this was not a trial about can patients with lower blood pressures be safely
managed on the ward and out of the ICU. Every patient in the study was
indeed managed in the ICU setting. So, I think that we need to be careful, right,
that it’s not extended beyond what was done. And in fact, you could even say that perhaps
these lower blood pressure targets were safe because they were so carefully
monitored within the ICU setting. We’re dealing with this acutely
right now in the United States. We’re doing a trial called Clovers through
the Petal Network that’s looking at sort of more vasopressors early or more fluids early
for patients with sepsis and septic shock. And this has been a big concerning point
that the patients who are randomized to the group that’s not getting early
vasopressors that they’d more likely go to the ward since they don’t have
the indication for being in the ICU. And at my hospital, we’ve pushed really hard to
say these patients are all the same at the time of randomization, and whether or not they
start on a little bit of norepinephrine, they should be carefully monitored in the ICU.>>Thank you, Terry. And that’s a nice perspective. Anders, from a continental European perspective?>>I’m not sure that, I think the opposite
may be more challenging because we come from a culture where we’re used to teach
and pray that hypotension is dangers. So the culture among our colleagues and
nurses is that the pressure has to be higher, and I think that’s why almost always in trials
on vasopressors see higher pressures measured and delivered than what the target is. That’s our culture. So, I think actually it will be more
difficult to adhere to permissive hypotension to change the culture from [inaudible]
hypotension that’s dangerous. Let’s be safe and put the
patient above the target.>>And I guess, and it kind of alludes to Paul’s
point as well, is what’s the trial told us about implementation of this evidence? So, the [inaudible] are very keen as to
public investment making a difference. So, we think we’ve probably
learned things from this trial. What should implementation
look like in our centers? Do we have key information from the trial
that would allow us to implement safely? Perhaps Cathy, can I ask
you that question actually? Implementation of your trial results.>>The implementation of trial results.>>Yeah, what have you learned in the trial
that’s going to be key for us to consider when we consider implementing
the results of this trial?>>Okay. So, the first thing I would
say is I sit with Anders on the notion that I think pooling these data with the
individual patient data from the other trials, so adding in another sort of eight, nine
hundred patient would be an important insight. I think that, if I talk about a perspective
of me being a patient, I’m not quite 65, but it feels like it’s getting reasonably
closer every year, I think that I wouldn’t want to be left drifting on vasopressors. I think one of the results for me that’s quite
interesting, and I know it’s a subgroup analysis and [inaudible] IPDMA is this
one on chronic hypertension. I’m not a clinician, as everybody knows, but
as I understand it, in that group the idea is to shoot for higher mean arterial pressure because that’s sort of their
usual normal sort of. So, I think that that sort of makes
me feel a little bit more comfortable. And so, I personally, when we looked at
kind of the data, what you realize is as people are being given these
vasopressor drugs, it’s not an absolute. It shoots above, you know, 65, but
you’re aiming between 60 and 65, and you could see that happening. The adherence in this study was fantastic, I mean the site when the patient
was randomized to the 60 the 65. So, for me, I think the implementation
is certainly one of understanding more from the data, both [inaudible] the IPDMA and
I think one of the nice things about, you know, you set up a trial, you got to bear in mind, we write that statistical analysis
plan, almost three or four years ago. And that’s what we forget
when we publish a trial. We go, that wasn’t very clever, they
didn’t think about X, Y, and Zed. You’ve got to think about when those sort
of bullets were sort of being put together, and then I think what happens is
you get important post-doc insight. And so, there’s a part of me that I agree
that dichotomania is not the way to go. On balance, I would like to be managed
and brought off vasopressors when sort of my MAP is between sort of 60 and 65. Does that make sense? I hear what Paul is saying about 60,
and what was interesting was how few of the [inaudible] how in both arms
that patients dropped below 60. So, that’s my take-away. I want to understand more, which is
always the case for me in trials. I’m always left with begging questions. But if you really want me to make a patient
decision now, I would have sort of 65 on my chest [inaudible] or
my forehead or whatever.>>So, there’s just, I mean there’s
the methodology question here about dichotomist outcomes, but
just before we get to that point, the notion of implementing trial [inaudible].>>Do I think the sites are the people to tell you how it should be
implemented because they did it?>>Yeah, yeah, but methodologically,
this is a complex intervention. We wanted an efficient trial. But, you know, within our culture in
the UK, did you consider doing things like process evaluation to really understand it. Sorry, yeah?>>Paul, what we did do is every time
where it, so we defined nonadherence as three hours where the MAP was above 65.>>Yeah.>>And so we didn’t see a
reduction in vasopressor does or discontinuation of vasopressors. And what we did do is we followed every single,
I think, occurrence of this with the size and found out exactly why, what
happened, and there’s some quite clear, where the issues were on
any sort of nonadherence. And they were very much focused,
the main reasons were sort of staff, logistical issues really, rather than
actually necessarily specifically about a patient they felt should be
pushed in the high blood pressure. And it was things such as maybe the nurse
overnight didn’t know the patient was the study, you know, we put stickers
everywhere, [inaudible] people too. They didn’t necessarily know that patient
was meant to be managed to 60 to 65, but it was overnight, and maybe it was a
nurse who just wasn’t familiar with the study. But also, maybe they had other clinical
priorities such as this physio, the patient, other clinical priorities that meant that vasopressors weren’t
necessarily their first priority in managing the blood pressure
so it wasn’t the first priority. When I mentioned that, if you’re
bringing in to usual care would be, would have just gone further and further
rather than sort of necessarily trying to actively titrate the work, clinical
queries that came up, so things like worry about low urine output, so you
can imagine why it did come up. There was a number of times, I
think that was the main concern with the potential renal concerns [inaudible]
the results didn’t sort of really back that up. So, I think there we do have quite a lot of data
on sort of the potential [inaudible] adherence, which could feed into, you
know, that could be implemented.>>But it is also important to say the
adherence was absolutely fantastic.>>Yeah.>>I mean I can’t remember, is
it something like six percent–>>It was six percent at the time–>>Of the total hours on vasopressors,
i.e., where people could have not adhered. There was hours of nonadhering
[inaudible] type of thing. So, I think when you come to the, so along
with the process evaluation, it was funded, as you say, Paul, in the efficient design, so
the idea of it was could you clinically embed, use data linkage, and drive the trial rapidly
to what we expected a recruitment period to be, and as you can see, sort of shot
that down because of that engagement. But I actually think the information will come
from the sites in terms of if those other sites, not in the trial, the sites in
the trial, that implementation. And some of the feedback is around the
reinforcement, the simple, the simplicity. We laughed about we wanted 65 sites, we’re
treating for 65 weeks to a blood pressure target of 65, you know, in 65 [inaudible].>>Sixty-five thousand [inaudible].>>There was a, we didn’t do it
for 65,000 pounds, not quite. But 65 million would have been nice. But there was a sort of a symmetry
and a kind of prompting and that kind of type stuff, but the sites are the experts.>>Okay. Thank you. And we’ll perhaps come onto that with some
more questions from the audience shortly. I’d just like to loop in Andrew at the end. By the way, I’ve got Andrew’s
Twitter feed up here. So, I would strongly recommend following Andrew. Adahltshousephd. Andrew is prolific on Twitter
with his statistical knowledge, and I wondered if we could develop this
discussion further, the point that Anders made, around dichotomist trials
and hypothesis testing?>>Yeah, so the latter question that came
from Twitter, I think very nicely stemmed out of Anders presentation, and he did a really,
really nice job sort of addressing this issue about dichotomania, and I love that graphic that
you put up showing the sort of three percent in the middle was like the darkest shade,
and then kind of slightly lighter shades as you moved further out, because
the challenge that we have is that three is not for certain the effect size. It’s our best estimate of the effect size
given the data that we have and becomes sort of progressively less confident
as you get away from that three. So, it could be as high as perhaps the
reduction of seven percent in mortality or it could be perhaps a slight harm. One point that Anders made, which I thought was
really keen is that there’s sort of an interest in doing less perhaps, and you
might wonder whether the bar for a trial is a little bit different when
you’re talking about approving a new medication or approving a new procedure versus this, which
is sort of like a de-escalation, or you know, a do less type trial, and
whether really this is a bit more of a noninferiority question
than a superiority question. And we don’t have time, unfortunately, to
go into a full discussion of the merits of noninferiority designs and their
challenges relative to superiority design, but I do think it’s fair in the
interpretation stage to look at the data almost through a noninferiority lens as much as
through a superiority lens and ask yourself, did the permission hypotension arm really need
to conclusively prove itself as much better to warrant adoption or is it the sort of thing
that maybe if the balance of the data suggests that it’s probably better and
probably not harmful at the very least, then maybe this is something that at the
very least probably can be safely done and explored without increase in harm. There was one other point that Anders touched
on, which I would like to expand on even more, was the supporting sort of sensitivity
analysis and subgroup analysis, and if there’s one thing I really,
really hope I can do at this meeting, if I don’t get anything else out of
this meeting, I’ll have done some good, is that there seems to be a belief
that in clinical trials there’s no need to do any sort of variable adjustment. There’s no need for a regression analysis. And it is technically true that there
is no need for a regression analysis to have a valid randomized trial. But, if you specify some variables
beforehand that you know are likely to have a strong relationship with the outcome,
that can improve the efficiency of the trial, that can improve the accuracy of the effect
size estimate and the confidence interval. And I thought it was very interesting
in the primary results to see that. And this is the sort of thing that [inaudible]
statisticians [inaudible] have to explain this, that the unadjusted result just barely wasn’t
significant, sort of in favor of the therapy. But the adjusted result was. And a lot of times people ask me, you
know, what am I supposed to make of that? And I say, well the truth is, if the
variables that you’ve chosen to adjust for are well justified and
strongly associated with outcome, the adjusted result is probably the more
accurate result, and if you’re asking me which one of those two to
believe, it’s the adjusted result. Where that gets complicated is people
worry that you’re sitting there fishing for which variables you should adjust
for until you get the result you like. So, if there’s something you can take away
from this, I think moving forward in terms of statistical analysis plans and trial
interpretation, I think the key is that when you have the situation where you
might want to adjust for some variables, if you’re able to specify them ahead of
time, that way you sort of have confidence that the investigators didn’t sit there and
say, all right, what if we adjust for age, I don’t know, take age back out, what if
we adjust for this, take that back out. But the truth is, the adjusted result is
probably the result that’s more trustworthy in this setting.>>Thank you very much Andrew. And what’s interesting about the
noninferiority design and, you know, whether the data would support what size of a trial you would have needed
to have very tight [inaudible].>>Yes, I mean it certainly
wasn’t a noninferiority design. I mean you need to be really, really careful
in how we kind of articulate the findings. It wasn’t superiority [inaudible].>>Any things you’d like
to come back [inaudible].>>Well I just think, so I agree with Andrew,
so that’s an easy one to come back on. And I think, you know, one of the things
that struck me even randomizing, you know, 2600 patients, not a good measure of frailty. I accept activities of daily living, etc., but the usual care patients actually were
slightly frailer, if that makes sense, and that was one of, and all the adjustment
we did was, was it the other way around, Francois is telling me, I’m sorry. So, the, so you’re absolutely right, the
permissive hypotension, we’re frailer, that’s right, sorry, forgive me, I
am getting at the wrong way around. So, the things that [inaudible] you get
imbalanced even in quite a large number, okay, and then I agree entirely
that, you know, prespecifying and publishing a statistical analysis plan, you
know, which we adhered to absolutely [inaudible] as you often do for some of the analyses. Maybe one of the other things that moves on is
statistical methodology in trial, and you know, there are people who argue for a relative
risks against odds ratio and all of that kind of type stuff, and you want to please all the
people, particularly when they come from JAMA. It’s amazing you want to please them. And so, you know, and I think labeling things
as pre hoc and post doc is really important. But, you know, I think, you know, we
did the job that we did prespecified.>>Howard wants to come back on
that one, I can’t [inaudible].>>Sorry Howard.>>Okay. So, so I think we’ve got
a few more minutes just to tidy up. Are there any further questions now, we’ve
developed a discussion from the audience. Please, if you could get a
microphone to the front, thank you.>>Thank you. I’m [inaudible] from Amsterdam. I have a problem with the
basic assumption of what type, what blood pressure we should aim
for, because you give a figure, and I was trying to teach my
fellows that they should aim for the blood pressure that is enough. And just by putting down a figure, it might be
very good to help us, get these pointers out, is to make an incentive that people
use less catecholamines or inotropes. So, I think that is a very good goal, but
it’s a little bit of an unsophisticated goal because what we want is that people discover in an individual patient what
is enough as a blood pressure. The same was true was true for
a cardiac output or whatever. I’m not interested is what
is the cardiac output, I am only interested to know
whether there’s enough. You have the size of your coat
that is probably enough for you, but I guess that it’s not enough for me.>>Okay, so Francois, do you want to–>>[Inaudible] fair enough. Great point. I guess what you bring up is,
you know, surrogate endpoints, and to what endpoint should we be resuscitating. I’m on a daily basis in the ICU fairly certain
that whatever we did for a patient was not great when the patient dies and was probably
good enough when the patient survives. The problem is that it’s difficult to know
beforehand, and I would, and I feel your, I mean I’m just saying, when
I have that clinician hat, I have to make decisions based on what we have. The issue, and Paul Young
speaks about this quite well, the issue is I would challenge the
assertion that we can currently rely on what most people propose as
endpoints of perfusion or function or I would challenge the assertion that urine
output is a good marker of kidney health. I don’t think creatinine clearance
is a great maker of kidney health. ACE inhibitor in diabetic nephropathy reduces
clearance and it improves kidney health. And the same way I would challenge the
assertion that lactates are perfect, they are, lactate levels are modified by catecholamine
stimulation, and so, I think it’s easy to fall into this viscous circle where lactates are
high, we feel like we’re not doing enough, we increase the doses of whatever we’re giving. The direct effect of this is to increase
lactate, and then the patient dies, and we say died in spite of best care, where in fact our best care
might have played a role in it. So, completely agree with you in principle. The devil is in the details is that I don’t
have great ways of directly measuring flow or function or predicting what will happen. If I felt confident we do have those tools,
they would have been incorporated in the trial.>>Okay, we’ve go three micro–
oh, sorry, Terry. Did you want to come back on that? I apologize.>>Yeah, just a very quick comment. I agree, but at the same time I think there is
a risk of giving too much power to the clinician at the bedside to say I know what’s
good enough or I know what’s best. I would have guessed that patients with higher
blood pressure at baseline would do better with higher blood pressure here, and yet
these study results show the opposite. So, our gut sense at the bedside is often wrong,
and maybe we shouldn’t have too much power. [ Inaudible Comment ] Right, but enough. If your baseline MAP is 80, then maybe
we would say perhaps 80 should be enough when that’s probably the wrong answer.>>Okay, we’ve got three more microphones
out to draw us through a close. I might start with the first question here. Thank you. We’re working.>>Hello?>>Not so well, actually. I apologize.>>Try again.>>Shall I, shall I, try again. Shall we get you a new microphone, and
then I’ll take another question first. [ Inaudible Comment ] We really want to hear you. Okay. Right towards the back. I can’t quite see– [ Inaudible Comment ] Oh, is it Steve? Yep.>>Hi. Is this working? Yeah. I’m just thinking about
the adverse effects. Were you surprised at the lack of difference
in the two groups with adverse effects, often one of the reasons why we’re reluctant to give vasopressors obviously
is the risk of adverse effects. There didn’t seem to be any major
differences between the two groups.>>I was happy we didn’t see huge differences. We, Paul and Alvin and then the entire team
put a lot of energy in discussions with sites. This was not a blind trial, so we were concerned that there might have been a risk given
our priority concerns that, you know, low blood pressure might be harmful, that
there would be differential reporting, and I guess looking at those
numbers we were very happy that there didn’t seem to
be differential reporting. If we were surprised, I don’t know.>>But I think what we didn’t
see was acute effects from vasopressors causing adverse events, and I
think maybe actually when you look at the trend of the time, it’s slightly later when
there’s a slight mortality change. It wasn’t these, I think acute
adverse effects that were causing any of the mortality changes or differences.>>We took a question here, finally.>>Okay. My name is Maria. I work as an intensivist and trialist
in Stockholm, and I am interested in how the blood pressure
was measured in this trial.>>It was measured as it’s
measured every day in routine care. So, this was very much a pragmatic trial. So, in most instances, just
like it happens most every day, it was measured via an invasive arterial
catheter, and as it usually happens, it was mostly measured via arterial catheters
rather than a catheter somewhere else. But if, you know, as it would happen, I
guess, in Stockholm, just like it would happen where I work, if you lose all your catheters
and you’re still running vasopressors, then they use the blood pressure
value they had in both arms. This was done very pragmatical.>>Thank you, thank you, Francois. I’m afraid time is getting very short. We’ve got one final question, I think.>>Hello. Peter McQuillan
[phonetic] from Portsmouth. I want to follow up on that last question. Recognizing that [inaudible] are
beneficial because the errors occur in both arms, hopefully, to the same degree. If we’re measuring blood pressure,
I’m just worried about the discipline of measuring blood pressure
and what is the blood pressure? Yes, it’s arterial. Is it radial, brachial, femoral? They’re different and they
change from time to time. And therefore, the baseline variance
of our measurement is quite noisy. And [inaudible] a difference in blood
pressure from a very noisy signal, then it’s going to be quite difficult. So, we may therefore have
underestimated the error so involved and hence underpowered our trial as it were.>>So, there are additional
integer research questions to pose for sure, and a point well taken. That uncertainty about what the blood
pressure is exists and has existed forever, and every day we have discordances between
art line, cuff, you know, femoral line, and that happened before, and it still happens.>>And current clinical care is delivered
to those noisy, perhaps ill-measured, perhaps measured in different– you know. So, the trial, if you like, tried
to parody that, if that makes sense, or that’s the pragmatic sort of approach. I think, you know, there’s a lot of work
around explanatory trials and measurements and measurement error and that kind of
type stuff that’s all really important too. But this is the noisy sort of rough and ready
mean arterial pressures that you guys use day in and day out to deliver clinical care.>>But perhaps the next time we have a bit
more discipline about what it actually is and where we measure it and make sure that
the [inaudible] stays in the right place of the heart, for example, rather than
patients moving up and done and all that.>>Thanks Peter. We’ll have to call it to a conclusion now. So, thank you as trial centers. Thank you for the investigators, the
editorial, the comments, the discussion. It’s been great. One thing I’ve learned from this is that were in
great shape in critical care in the UK in terms of reaching out to our populations. I think the recruitment dates across the
24-hour cycle is fabulous and the NIHR are going to be delighted that we are
reaching patients with research. So, thank you very much.

Leave a Reply

Your email address will not be published. Required fields are marked *