The ProMISe Study: EGDT RIP?

The (1)
Today is another landmark in the history of sepsis research. The long awaited results of the ProMISe trial are finally published in the New England Journal of Medicine. At 56 hospitals across the United Kingdom we’ve been working tirelessly to recruit patients to this awesome, cutting edge randomised controlled trial comparing early goal directed therapy to standard care for patients with severe sepsis.

This last year has been amazing for research into early goal directed therapy. We’ve been lucky enough to have two huge randomised controlled trials published from Australasia and the United States. The ARISE and ProCESS trials both found no difference in 90-day/60-day mortality between patients who were randomised to receive early goal directed therapy and those who received standard care. But would things be different in the United Kingdom? Could there be a small treatment effect that was not detected in the previous two randomised controlled trials? Despite their large sample sizes the 95% confidence intervals from those trials could not exclude the possibility of a 20% relative risk reduction for 90 day mortality. If there really is a treatment effect of that magnitude then early goal directed therapy could still be useful.

Who was included in the ProMISe study?

Essentially we had patients with septic shock. All patients had to have two SIRS criteria, a known or suspected infection and either hypotension (systolic BP less than 90 after a litre of fluids) or a lactate of at least 4 mmol per litre. After consent (or assent from a relative or independent representative), the patients were randomised to receive either standard care (in which case the clinicians could do whatever they felt best) or early goal directed therapy (which involved following the algorithm below). Because of the nature of the intervention, all patients in the intervention group ought to have had a central line and an arterial line inserted, and should have undergone 6 hours of continuous ScVO2 monitoring.

ProMISe flow diagram

All of the patients in this trial were then followed up for the primary outcome of all-cause mortality after 90 days. They looked at other outcomes too, such as organ dysfunction, length of stay and they ran an economic analysis. Ultimately, this was a large trial: 1,251 patients took part, which gave the study enough power to detect a 20% relative reduction in mortality assuming that the 90-day mortality would be (a whopping) 40%.

What were the key findings of this study?

The bottom line is that there was no difference in 90-day mortality, so the findings of this trial don’t support the protocolised care used in this trial. In total, 29.5% of patients in the EGDT group and 29.2% in the control group died. The 95% confidence intervals were fairly tight too. The relative risk was 1.01 with confidence intervals from 0.85 to 1.20.

The secondary outcomes tell a similar picture: costs were similar, length of hospital stay was similar. Patients in the EGDT group had significantly higher SOFA scores at 6 hours, received more cardiovascular support and had longer stays in the ICU than patients in the control group.

So, for sure, we can conclude that ProMIse showed no difference between EGDT and standard care, which was also the finding in the ARISE and ProCESS trials.

But were there any methodological problems with this trial?

This is a large, well conducted randomised controlled trial at 56 centres. The groups were well matched at baseline, data were analysed by the intention to treat principle and there was a very good follow up rate for the primary outcome. In general, the trial holds up to scrutiny and deserves to be published in the NEJM as it’s definitely a landmark trial. It definitively shows that rigidly following that protocol in patients who meet the criteria for septic shock isn’t beneficial compared to what we’re doing right now. There are still a few questions we might ask, though…

Is this really EGDT RIP?

Possibly. And possibly not. In this study 92.1% of patients in the intervention group had a central line inserted within 6h. But, in the control group, 50.9% of patients also had a central line inserted – and the time to insertion was similar in both groups. In the intervention group 74.2% of patients had an arterial line inserted, but so did 62.2% in the control group – and again the timing was similar. Both groups received fluids in the first 6h (1.75L in the intervention group, 1.5L in the control group), both groups received vasopressors (53.3% versus 46.6%). Both groups were admitted to ICU (88.2% versus 74.6%). More patients in the intervention group received dobutamine and packed red cells but numbers were small, so if those interventions really are the beneficial ones then this study is probably underpowered to detect differences in that subgroup.

Perhaps what we’re seeing is contamination, where patients in the control group are actually exposed to the intervention being studied – or at least parts of it. Because EGDT is so ingrained in us and because our main focus is to treat patients, maybe the sickest patients in the control group were still receiving EGDT. And maybe clinicians selectively using EGDT in the sickest patients from the control group is what made this a negative trial.

What’s more, the inclusion criteria seem to incorporate a fairly broad group of patients. We have the hypotensive patients, who I think we would all agree have septic shock. But we also have the normotensive patients with a high lactate – which could be as little as 4.1 mmmol/L. If those patients cleared that lactate within an hour, returned to normal physiology and received early antibiotics (which all patients did) would we really want to insert central lines, arterial lines and continuously monitor ScvO2 for 6 hours? I’m not so sure that we would in our current practice.

To my mind, this trial justifies our current approach. If your septic patient is sick, go for the EGDT approach. Insert a central line, insert an arterial line, fill the patient up, give vasopressors if they’re still hypoperfused. That’s what happened in this trial – it doesn’t knock that approach. It they’re not so sick, then don’t.

Maybe that’s all this trial is really telling us. And maybe that’s a limitation of the RCT to answer these questions.

Where next for sepsis care?

With three huge RCTs just published, we’re now waiting to see what will happen when the data are combined and subgroups can be analysed with more statistical power. Is there any effect of EGDT in the hypotensive subgroups? Are there any groups that might benefit? Are there any particular elements of the pathway that seem to produce an effect?

ProMISe hasn’t killed EGDT. Not yet. There are still a good few questions to answer.

Rick

 

35 Comments

  1. Jonathon Hurley

    Could someone explain the powering of this study? This is something I’m still getting my head round. It was powered to detect a 20% relative risk reduction based on an assumed mortality rate of 40% – does this mean that it should detect a mortality of 32% or less in the EGDT group compared to the standard care group?

    Thanks!

    Reply
    1. richardbody (Post author)

      Hi Jonathon,

      Yes, that’s basically it. They considered that the usual care group would have a mortality of 40% and that the minimum clinically significant difference was a relative risk reduction of 20% (equating to an absolute risk reduction of 8%). In case you’re interested in the detail, here’s my take on reproducing their sample size calculation:

      The authors wanted to compare independent proportions – i.e. the proportion in each group that died at 90 days. They thought that 40% of the control group would die and wanted to power the study to detect a difference of 8% – i.e. 32% mortality in the intervention group. They wanted 80% power and a two-sided alpha of 0.05, which is fairly standard of course.

      The number of patients we need in each group in this case is given by:

      n = f(a,b) * ((p1(1-p1)+(p2(1-p2))/((p1-p2)(p1-p2))

      Here’s a reference for that… https://www.rnoh.nhs.uk/sites/default/files/sample_size_formula_for_a_comparative_study.pdf

      f(a,b) is a function of the required alpha (a, which is 0.05) and beta (b, which is the power – 0.8) and comes to 7.85.

      p1 is the expected proportion in the intervention group (0.40) and p2 is the expected proportion in the control group (0.32). With those assumptions, we need 561 patients in each group or 1,122 patients overall. The authors thought they’d have a 6% loss to follow up, so they multiplied 1,122 by (1 – 0.06). That gave me a total of 1,190 patients. I presume there’s a rounding error or minor difference in the formula because the authors came to a similar but slightly higher figure of 1,260 patients.

      That’s how the sample size is usually calculated to compare independent proportions though. (DOI, Coming from a non-statistician!)

      Rick

      Reply
  2. Matt Astin (@mastinmd)

    Excellent review! I agree that this is not the death bell for EGDT. To me, this is more ammunition for “the right treatment for the right patient at the right time.” As pointed out, there are some patients who require the interventions from EGDT, but others that do not. The same can be said for other interventions we have studied in emergency medicine (tPA for stroke, TTM at 33 degrees post cardiac arrest). This just adds to our armamentarium for treating sick patients.

    Reply
  3. Shariq

    As always an amazingly helpful blog Rick. Have to say though that these three studies shouldn’t make us complacent. There should always be a sense of urgency when managing this slippery group of patients.

    Reply
  4. emcrit

    Rick,

    Fantastic summary! Would quibble with one comment:
    “If your septic patient is sick, go for the EGDT approach. Insert a central line, insert an arterial line, fill the patient up, give vasopressors if they’re still hypoperfused.”
    That’s not the EGDT approach, that is just good critical care and was actually the control group of EGDT.

    What ProMISE and the other two showed is that CVP and ScvO2 are not necessary values in order to manage a comprehensive resuscitation. That is what these trials put in the grave.

    Reply
    1. matt anderson

      Also shows hat we shouldn’t be giving fluids based on a CVP, pRBCs and dobutamine based on a number alone and overall this is a protocol that dr rivers made, used and showed benefit. For hospitals using this protocol, this is evidence that a specific protocol is not needed nor should be considered a guideline or SoC. As Weinberg says ‘you don’t have to do a lot of s*** just give a s***’.

      Reply
    2. Simon Carley

      I think Scott is right here. That;s what we do with the sick septic patients.

      Rick did you mean put the Central line in as you are heading for vasopressors rather than for CVP monitoring? I suspect you did as that’s what we do (in Virchester) rather than chase a CVP.

      S

      Reply
      1. richardbody (Post author)

        Absolutely, I wouldn’t blindly chase a figure with an AUC of just over 0.5. That’d be like using risk factors to decide which patients have ACS! 🙂 Having said that, you will I’m sure tear your clothes and beat your breasts when I admit that I would still monitor CVP in someone in whom I’m worried about fluid balance. (I still ask about cardiac risk factors in patients with chest pain, too) It’s another piece of information to inform you. As clinicians, our job is then to decide how useful or useless that information is. With CVP, it’s not just the absolute figure that’s helpful and I wouldn’t rely on it – I’d use other parameters to assess filling, of course.

        Lastly, it’s important to acknowledge I’m not the expert in this area. CVP is something I trained with and, like many people, coming to rely on alternative parameters involves a learning curve. As practising clinicians we’re constantly finding out that dogma we’ve previously clung to is wrong. With every paradigm shift there’s a time of adjustment. Having recently finished the ProMISe trial (where we did chase CVPs) and, before that, having used CVP to guide filling for a long time, I find myself in something of a period of transition when it comes to assessing filling in sepsis. I imagine the whole field is in that sort of state right now. #FOAMed is great to help with that process – especially sites like EMCrit and RebelEM!

        Rick

  5. Wojciech

    I totally agree with your concerns. How much of “standard” therapy is really EGDT? When I read the ARISE trial, it came to me a similar doubt. Many of interventions in the control group in the three trials mentioned, including pharmacological and other strategies, have been learned through many years and applied consciously or unconsciously, producing what you correctly call “contamination”.
    In the other hand, a septic patient is always a critical ill one, therefore, he or she is always sick.

    Reply
    1. Matt Anderson

      Didn’t the arise group publish data before that showed there was minimal influx of rivers EGDT into practice?

      Reply
  6. Pingback: Septicémie, sepsis grave et choc septique | thoracotomie

  7. Matt Anderson

    I would argue that this does in fact put a nail in the coffin in RIVERS EGDT. PROCESS, ARISE and PROMISE show we don’t have to follow his protocol as the SSC pretty much recommends. However, in my mind EGDT is still useful–choose a marker (map, lactate, even if u want it to be intermittent ScvO2 or some other marker of perfusion) and shoot for a goal, if you aren’t getting there or the patient is getting worse — something isn’t right — wrong abx, lack of source control, not septic shock. Rivers study was a strict protocol — promise puts the nail in the coffin that his protocol is not as good as what we are doing now.

    Reply
  8. Niklas J

    Thank you for a great summary! I would like to comment on one thing; patients in septic shock have low blood pressure and hyperlactatemia but we don’t really know if they are hypoperfused.

    Reply
  9. Rob Mac Sweeney

    While high quality critical care research is always welcome, it seems to me contemporary practice has abandoned the unscientific goals of Rivers long ago. Has anyone targetted a CVP of 10 in the last decade, or transfused to 10 because the ScvO2 is low? The only surprise has been EGDT hasn’t demonstrated the degree of harm that the individual goals would suggest.

    Reply
    1. srrezaie

      Hello Rob and Rick,
      Yes Rob, I am on your bandwagon…..what the trilogy of sepsis studies has shown us is that there is no difference between EGDT vs “Usual” care….not that EGDT causes more harm. The studies have shown us early IVF, early abx, early source control the so called ABCs of resuscitation the use of EGDT, Protocolized care, or “Usual” care doesn’t matter as much. In other words don’t stop using IVF, CVCs, and A-lines when needed, but instead stop following CVP and SVO2 to guide management.

      Rick for you, didn’t ProCESS look at 60 day mortality and ARISE and ProMISe looked at 90 day mortality? Small point, but I understood what you meant. Great summary and interpretation as always.

      Salim

      Reply
      1. richardbody (Post author)

        Thanks, Salim! I’ve just replied to you, Scott and Simon in one – above. I’ve corrected the error you mentioned.

        Great post on REBEL EM, by the way!

        Rick

  10. Pingback: The Protocolised Management in Sepsis (ProMISe) Trial - R.E.B.E.L. EM - Emergency Medicine Blog

  11. Pingback: EGDT IS DEAD, LONG LIVE EGDT |

  12. DocXology

    ‘Usual care’ can represent a wide spectrum of practise whereas a protocol standardises that care. Therefore cookbook medicine results in a central tendency to the mean. It lifts the standard of low performers and it can potentially compromise the outcomes or highly resourced/high functioning system.

    Therefore, PROCESS/ARISE/PROMISE all tell me one thing. If you have little experience with managing severe sepsis then you can still mimic the outcomes of these trials.

    I would be greatly interested to see the results of such a trial in purely regional and remote centres where the impact of EGDT could make a huge difference..

    Reply
  13. Pingback: LITFL Review 174 - LITFL

  14. Pingback: EGDT, it had such PROMISE

  15. Pingback: FOAM Eye-Catchers 9: Early Goal Directed Therapy is Dead | Emergucate

  16. Pingback: SGEM#113: EGDT – ProMISe(s) ProMISe(s) | The Skeptics Guide to Emergency Medicine

  17. Pingback: SGEM#113: EGDT – ProMISe(s) ProMISe(s) | The Skeptics Guide to Emergency Medicine

  18. Pingback: » Leerstof Onderwijs SEG 7 april Spoedeisende Geneeskunde WFG

  19. Tim Coats

    Interesting discussion on ProMISe – a couple of extra thoughts. The things that were different between the control and intervention groups were; (1) a strict numbers based protocol to follow, (2) continous ScvO2 measurement for all. The results showed that the addition of theses two factors did not give improvement in survival. Is this “testing EGDT”? As EGDT has many components I don’t think that ProMISe really “tested EGDT”. Patients in both groups still got early indentification, early fluids, early antibiotics, critical care review, and treatment titrated against individualised targets (ie those set by the treating clinicians rather than by a protocol). So ‘good sepsis care’ still includes components of EGDT – but we now have a better idea of what is important and what is not.

    Lastly, the control group gives us a good benchmark of what is going on in UK centres who are interested enough in sepsis management to take part in the trial. The control group interventions can therefore be used as a benchmark for future audits and can be a good way for you to assess your sepsis care locally (amount of fluids, type and frequency of interventions etc).

    Future directions for sepsis research are probably in looking at the dynamic processes rather than just having a single point in time definition of sepsis, with better methods of assessing changes in cardiac and endothelial function in response to treatment. Maybe the legacy of ARISE, ProCESS and ProMISe will be that we move from a ‘recipe’ to the more personalised approach of the dynamic assessment of individual patients.

    Reply
    1. richardbody

      Thanks a lot for taking the time to comment, Tim. It’s fantastic to have your insight as an author of the NEJM publication. If that is the legacy of ProMISe, ProCESS and ARISE it sounds very positive. That’s a great way of thinking about the bottom line of these three trials. They do strongly suggest that having a single one size fits all ‘recipe’ for managing sepsis isn’t the way forward.

      Running the study locally also highlighted to me the size of the challenge we have *recognising* sepsis. It will be great to hear more about ProMISe from Paul Mouncey, who’ll be speaking at RCEM Manchester 2015 (September 28-30) and at EuSEM 2015 in Torino (October 10-14). In Manchester we’ll also hear from Ron Daniels, Chief Executive of the UK Sepsis Trust. I’m very much looking forward to the further discussion that ProMISe will stimulate!

      Rick

      Reply
      1. lawrencelynn

        Rick
        Excellent thoughts. I enjoyed it.

        Is not it true though that the calculation of the number to treat assumes a common condition under test?. There is no unified phenotype of sepsis. For this reason none of these studies will reproduce positive results of another except as a function of chance.

        Thete is another problem which is just as severe as the lack of a single condition under test. This is the use of static simple threshold targets. We warned of mathematical embellishment in our work in the 1990s when the treatment goals for sepsis were a high DO2 driven by Dobutamine, high volume NS, and Hgb. This was the 1980s Era of supra-normal DO2 targets in the treatment of sepsis using the PA cath. Rivers modified those efforts (which had failed two decades earlier) by lowering the targets and using those obtainable from the ScvO2 catheter. (SVCO2 and CVP instead of SVO2, PCWP and DO2 of the 1980s).
        .
        Since we had both CVP and PCWP in the 1980s we knew CVP was unreliable. Of course the choice of thresholds like CVP of 10 and Hgb of 10 comes from 1980s thinking. Those are the numbers of fingers on your hands and if you think that is a coincidence then I can show you that the number 10, 100, and 100-10 were the ritual care thresholds of choice in the late 20th century critical care including the tidal volume of 10cc/kg, 10 bands, 100-10 HR and 100-10 SBP. (100 is the number of soldiers under a centurion but has no physiologic significance. These 1980s thresholds spilled over into ARISE, Process, and Promise. They were always capricious at best and in fact more correctly defined as silly. We told everyone who would listen (which was virtually no one) that they were artifacts.

        Threshold science does not work for complex dynamic conditions. That the static guessed threshold targets based on the number of fingers on the researchers hands exposes the capricious nature of 20th century threshold science but the science would not work if any other number was chosen.

        These brief axioms in this editorial are easily understood and they show why we cannot think in static threshold terms applying p values and AUROC to data and time fragments.

        http://www.ncbi.nlm.nih.gov/pubmed/24834126

        Many of you know we predicted this but how did we know. These 4 axioms made it clear that any threshold is an artifact especially they being derived from human anatomy are used to define pathophysiology.

        All of you have summed to up very well.

        Know the pathophysiology.
        Know the dynamic relational patterns
        Spend the time at the bedside to learn the value and pitfalls of volume assessment tools.
        Don’t let a patient with lactic acidosis die with an empty tank and a vasopressor generated false sense of blood pressure security.

        I look forward to your criticism/ comments

      2. DocXology

        The anti-EGDT brigade argue that sepsis is a heterogenous entity that requires nuanced, tailored care that defies a ‘one-size-fits’all’ approach. Indeed even the premises behind each intervention within the EGDT protocols have shown to be individually flawed.

        Therefore, it is surprising that three mega-trials were unable to demonstrate clear superiority of expert, individualised care compared to a cookbook approach to sepsis. Thus the implication is that EGDT must work for the majority of septic patients with only a small fraction that could be shown to be harmed by such an approach.

        Perhaps this is the uncomfortable truth, for most of the time a trained monkey could do our job. Only in the rare circumstance does expertise influence the outcome.

        Nevertheless, I guess we can keep our jobs because it may impossible to predict which ones they might be.

  20. Pingback: Sepsis Guidelines: EGDT without the "G"?

  21. Pingback: Research and Reviews in the Fastlane 076 - LITFL

  22. Pingback: A Prof. on PROMISE | The Rolobot Rambles

  23. Pingback: JC #9 – Djupdykning i septisk chock – Om protokollbaserad handläggning | internmedicin

  24. Pingback: LITFL R&R - Best of 2015 | LITFL: Life in the Fast Lane Medical Blog

  25. Pingback: Sepsis Trilogie deel II – Nieuwe studies, nieuwe inzichten | fan of EM

Thanks so much for following. Viva la #FOAMed

Translate »