Possibly the first thing to consider is why there are 3 studies in the first place? You might want to think and comment about why that may be. The given rationales were undoubtedly retrospective (and therefore not the real reason), given that the study dates overlap so much (ProCESS Mar 08-May 13, ProMISE Feb 11 – Jul 14, ARISE Oct 08- Apr 14), with protocols written and finding obtained before the results of any of the others were known.
In any case, the authors give their rationale as:
1 – ProCESS were first to publish, they needed no rationale other than to answer the question.
2 – ARISE was to see whether results of ProCESS could be replicated outside the US, and in ‘non-academic centres’ (I do wonder how the ‘non-academic’ centres felt being approached to take part in the study?!).
3 – ProMISe was designed to also look at cost effectiveness, and rather disappointingly to investigate an area where “the reported mortality for septic shock is high” – just when we were starting to feel good about ourselves after the Olympics! The introduction to the paper also makes another argument to do with possible false negatives in the other two. The problem comes from a lower than expected mortality in the control group, and is something I think we’ve discussed before.
In the ARISE study for example, the investigators ‘powered’ the study at 80% to look for a 20% relative risk reduction, meaning the chance of a type 2 error (not seeing a difference if one exists) would be 20% (because the chance of a type 2 error = 100-power). Put another way, if there was a 20% relative risk reduction, there would be an 80% chance of finding it. 80% is a standard power, it is accepted that medical research would rather miss a difference than find one where it doesn’t exist (so we accept a 5% chance of a type 1 error, but a 20% chance of a type 2). The problem in the ARISE study was that the recruitment target (the numbers needed for 80% power) were based on a higher than observed mortality. They expected a 38% mortality in the control group, but got 18.8%. Why is this important? If the mortality is 30.4%, a 20% RRR is a 7.6% absolute risk reduction. If the mortality is 18.8%, a 20% RRR is a 3.76% ARR, a smaller difference. To detect this smaller difference, it is intuitive that larger numbers are needed. Therefore, if the observed mortality is lower than expected, more participants would be required for that power, meaning the power is lower, so the chance of a type 2 error is higher, and so the resulting negative trial may actually have been positive. It’s the same story for ProCESS, where the observed mortality was 18.9% against a predicted 30-46%. Unfortunately, in the ProMISE trial they went on to make the same ‘mistake’ – with an observed baseline mortality of 29% rather than the anticipated 40%! (although in the discussion it seems that when they did it, it was unlikely to make a difference!) They based the 40% on ICNARC data so either there’s a Hawthorne effect, ICNARC data doesn’t apply to participating centres, suddenly the UK isn’t so bad, or there’s a problem with ICNARC data.
I’m not at all surprised by the results, usual care is EGDT done pragmatically, leaving out the bits that we don’t believe make a difference and tailored to the individual patient. To my mind that is the interesting part of these studies, they show us what ‘usual care’ is. What is seems to show is that usual care, when considering interventions rather than monitoring, is not that different from EGDT. For example, volumes of fluid given may have been statistically significantly different, but does anyone believe that the 250ml difference in ProMISE, or the 500ml difference in ProCESS is of any real meaning? Dobutamine administration and red-cell transfusion seem to be where there is a difference, but I’m not sure that even if they were both shown to be the panacea for sepsis management we would embrace their widespread adoption (after all, we didn’t post - Rivers). 30% of the patients in each arm of ProMISE group received approx. 1 litre of colloid, clearly the message regarding harm with colloids is still not quite out there.
All studies used a 6hr intervention and looked for a 60 or 90 day mortality as their primary outcome. Do you think that is appropriate? I can see how being alive at 60-90 days is what the patient and society would view as a meaningful outcome, but is it reasonable to think we can affect 90 day mortality in just 6 hours? Do the results mean anything when we don’t match everything that happens over the next 89 days and 18 hours? Should we first investigate whether there is a causal link between physiology at 6 hours and mortality at 90 days? It potentially gets even more tenuous when you consider that the time to randomisation (and therefore different treatment) was approx. 3 hours in all studies, half of the intervention time. If these 6 hours have a 90 day effect I’m pretty sure the proceeding 3 hours will do too.
I also think any differences in outcomes must also be considered in the setting of a trial intervention. All papers make mention of likelihood to be admitted to ICU – if we were recruiting to this trial, all enrolled patients in the EGDT group would come to ICCU as that’s the only place they could receive the intervention so it’s not necessarily a marker of ‘sickness’ or of successful resuscitation in the ED.
With regard the economic evaluation in ProMISE, I don’t really have any understanding of QUALYs, although it looks like no significant difference – there’s a link to some info here. I’m also not sure about the cost data, but given the standard deviation for cost in both groups is so large, I wonder if the study was not large enough to give meaningful data? Hopefully someone will comment about this.
Undoubtedly, sepsis is an illness with a high mortality, and many associated deaths are avoidable. These studies don’t tell us to be any more relaxed about sepsis, rather we should be focusing our efforts on those interventions that seem to be effective now we can be increasingly justified in not being distracted by those that don’t. In the ARISE study, the mean time to first antibiotic was 70 minutes from presentation. The ProMISE study makes no mention either in the paper or in the supplementary appendix (strangely), but I can’t believe we’re any better in the UK. Would reducing this time have any effect? How would we do so? As discussed when looking at ProMISE, focusing on getting the basics right would probably be the best intervention we could make.
The surviving sepsis campaign have changed their 6 hour bundle because of these papers, you can find the new bundle here. I don’t think this will end the campaign by any means, but the next set of guidelines may be radically different. Although Rivers’ results have not been replicated, he has altered ‘usual care’ such that landmark sepsis trials are overestimating baseline mortality by 10% or more, and has taught the world to treat sepsis aggressively. Still to my mind one of the two most important papers in critical care medicine.
There’s been a lot written about these papers (inc. but not limited to: St Emlyn’s, The Bottom Line, R.E.B.E.L EM), and I’d encourage you to have a look around, but ultimately you need to decide what difference, if any, they will make to your practice. Over to you…