JC: The beauty of simplicity. ANDROMEDA-SHOCK St Emlyn’s

For many years, we have enjoyed discussing the concept of euboxia in critical care. If something has a normal range, and abnormality is bad, surely using medical interventions to restore it to the norm can only be helpful? Lactate has been fed, saddled and ridden all the way to the races on this. It remains a key component of international screening metrics and clinical guidelines. There are opposers to this logic and we certainly believe at St Emlyns that lactate is often misunderstood and poorly applied as a point of care test. Make of this what you will. Either way, a lot of scientific attention has been paid to the baseline level and the speed and success of lactate clearance, as a proxy marker of successful resuscitation

Do we need this? What about good old fashioned clinical assessment of shock resolution? Shock windows were described many years ago and feature heavily in textbooks;  the skin is one such window, and capillary refill time (CRT) following direct pressure is taught on most life support courses as a central tenet of shock assessment. However, it is subjective, operator dependent and potentially subject to bias. Not like the good old objective number spat out by whatever flavour of gas machine you invest in at your ‘shop’.

How do the 2 compare? Well, helpfully, JAMA proposes to have the answer for us. A recent ambitious study seeking to compare objective numerical values to a subjective clinician dependent interpretation, to guide resuscitation during early septic shock. Interesting stuff and hats off to the Latin America Intensive Care Network for completing and publishing this work.

What did they do?

The alternative hypothesis for the study, was that resuscitation guided by peripheral perfusion would be superior to that guided by lactate level targeted resuscitation in patients with early septic shock. The authors conducted a large randomised controlled trial to answer this question.

Are your eyebrows raised already? If not, they should be. Why on earth do we think capillary refill time (CRT) would be superior to lactate clearance?

How did they do it?

This was an open label, multicentre RCT across 28 South American Centres. The authors convened a Trial Steering Committee and a Data Monitoring and Events Committee which is good practice for a study of this size. The protocol and analysis plan were pre registered and published, which is all good stuff as well. Central randomisation via permuted block at patient level was used, which should have maintained allocation concealment; an important factor in the context of an open label trial.  

This study was set in the ICU which has several important implications I think. Firstly, this affects the generalisabilty of the findings to other acute care areas. Secondly, would you use CRT on your unit when you have access to arterial line monitoring, central venous pressure, focussed echo etc? Would it be drowned out by other information, or if these study results were positive, would you be happy to go back to basics? I’m not sure I would (Ed – but in many health economies where invasive monitoring is limited this strategy might well be a realistic alternative).

The inclusion criteria seem straightforward – suspected infection, lactate >2, vasopressors needed for a MAP 65 following 20ml/kg fluid. We can debate the details, but I think most of us would agree this group to be a pragmatic representation of septic shock. Exclusion criteria were pretty standard also; bleeding and DNR orders, although I was a bit surprised to see severe ARDS in here.  

What about the intervention?

Following randomisation, there was an 8 hour intervention period. Patients were assigned to one of the following:

  • CRT – pressure was applied to the right index finger with a glass microscope slide, maintained for 10 seconds after blanching. The refill time was assessed using a chronometer. >3s was defined as abnormal. This was assessed every 30 minutes until normalisation, then every hour.  I assume a chronometer is a watch/clock. I cannot see why they have used extra letters here, other than to befuddle me with long words.
  • Lactate blood levels, assessed every 2 hours. A 20% decrease was considered to be evidence of effective resuscitation, with an aim to normalise.  

Abnormal results from either strategy triggered an interesting cascade of intervention, on top of baseline critical care for sepsis. Assessment of fluid responsiveness, tailored fluid therapy, trial of a higher MAP target and inodilator use. The supplements describe baseline and increasing levels of care.  

It is undoubtedly a strength of the study that the authors tried to structure the response to inadequate resuscitation. However, they also allowed local variation at the discretion of treating physicians and surviving sepsis campaign guidelines were also endorsed. The latter have received a lot of recent criticism as highlighted above. These issues have the potential to introduce bias and affect validity.

And the outcomes?

The authors chose a primary outcome of mortality at 28 days. Follow up by phone or death register. There were lots of secondary outcomes.

The sample size calculation suggested a need for 420 participants. This provided 90% power to detect a 15% ARR with an alpha of 0.05. I can’t resist getting into this right now, and I don’t think it would be going too far to suggest that this ARR is crazy. A NNT of 7 for CRT guided resus, compared to lactate guided? A relative risk reduction of 50%?  Better than the NNT for appropriate and early empirical antimicrobial therapy? I have no idea how they made this leap – although the cynic in me would suggest that they did the sample size samba – it was a South American study after all….

It is also worth pointing out again here that the authors determined CRT as their intervention, with lactate as the control. This was also a superiority trial. I have no idea why they were so confident about all this. And I am not sure why this was not a non-inferiority trial; I would have been very interested to see simply that CRT guided resuscitation is no worse than lactate guided. A genuine result on this could still save a lot of blood gases.  The DMEC also had no stopping rules I can see, although I appreciate many study groups don’t publish these in the central manuscript.

The Results

They cracked this in a year, recruiting and randomising  424 participants, which is excellent work. However, the consort diagram raises a few early concerns. 68% of assessed patients were excluded during this year, with 102 deemed to be eligible but not enrolled. The excuses here are pretty thin – logistic issues, physician preference, lack of consent etc…. 102 is >just shy of a quarter of the trial. In essence, this makes the data look more like a convenience sample, which introduces further risks of bias.

No patients were lost to follow up and all were included in the ITT analysis, which is good, but perhaps represents the follow up methods of death register and telephone follow up, which are not great. Baseline characteristics were mostly comparable between intervention groups suggesting success of the randomisation process, although a gender difference of 8% is quite high. The mean APACHE 2 score was 22 in both groups – much higher than the mean of 18 in the PROMISE trial for example, showing how international sepsis trials can struggle for generalisability.

There were protocol deviations in over 10% patients. Mostly around deviation from the suggested resuscitation algorithim I think. This raises some questions of internal validity. And there are questions already. Questions on questions….

And the primary outcome?

For the key result of death within 28 days, results were as follows: 34.9% (CRT guided) vs 43.4% (lactate guided). A Hazard ratio for death of 0.75, but with the 95% CI crossing 1. A p value of 0.06. Close, but no cigar.

Or is there a cigar? Peripheral perfusion is a cheap reliable and easy method of assessment in septic shock. Although these results don’t prove it is significantly better than lactate clearance, the likelihood of it being as good if not better is very high indeed. Had these results been statistically significant, we would be talking about an absolute risk reduction of 8.5%, with a NNT of 12. Just think about what this result suggests – to save one life, you would need to abandon a resuscitation strategy guided by lactate clearance and rely on one guided principally by peripheral perfusion assessment in only 12 patients. What’s not to like about that?

There were some interesting secondary outcomes. Significantly less organ dysfunction at 72h and less fluid administration were highlighted in the CRT group. However, these variables are more continuous; as such it is often easier to show a statistical difference but the question of clinical significance remains. The SOFA score was reduced at 72 hours by 1 and the fluid admin by 408mls mean. I am not sure I buy this as clinically remarkable.

In the prespecified subgroup analysis, patients with a baseline SOFA score of <10 seemed to do significantly better with peripheral perfusion. This is quite interesting. Why would a lack of reliance on lactate clearance in the less sick lead to better outcomes? Perhaps because you are not then forced to resuscitate as aggressively? Perhaps you give less fluids? These are uncomfortable conclusions that make us try and face up to the fact that our sepsis care can often be overzealous, to the later detriment of the patient.  Hypothesis generating only, but food for thought. Once the SOFA score is >10 the HR levelled off at 0.98 implying no difference in outcome between intervention groups.

All very interesting. Any concerns with the design or results?

Unfortunately, there are quite a few. This doesn’t negate the value of information provided by the study. But some common issues are worth digging into.

The failure to achieve statistical significance could well be a type 2 error here. Had they powered looking for half the proposed ARR, they would have required 1580 patients; a study of this size could well have led to a more convincing result. Have a go with a sample size calculator – you will be amazed at the margins that lead to far higher recruitment figures.

There are lots of questions on generalisability – is Argentinian sepsis the same as that in the UK? Is ICU sepsis the same as ED sepsis? Mortality rates were quite high with both interventions I thought; the mortality in the recent PROMISE study was 29% in both groups for example. As highlighted previously, the baseline APACHE 2 was lower in PROMISE so this outcome may simply reflect the severity of illness.

I still have questions about the CRT intervention – what about those with peripheral poor perfusion, known vasculopathy, Raynauds disease etc… Can this intervention really be standardised? I would also ask questions about the timing of review – as well as a different method of assessment, this study also randomised assessment to 30 minutes (CRT group) and 2 hour intervals (lactate group). Although you may tell an intensivist to just check the CRT at 30 minute intervals, you will not stop her sneaking a cheeky look at the latest gas, UOP, CVP, art line trace, HR etc…. This type of repeated frequent clinical assessment can only ever be beneficial. Perhaps this is one of the lessons from the study; that providing you regularly reassess, simple assessment measures are just as good if not better than objective numerical goals.

As discussed, the number of exclusions and protocol deviations make this more of a convenience than a consecutive sample. As such there is a decent risk of bias in the outcome.  The study was also open label and as such introduces conscious and subconscious bias. The authors tried to reduce this risk through a hard outcome measure and standardised protocol, but also allowed clinician preference on several measures.

What else is interesting?

Trials of hypertensive therapy achieved resuscitation targets 40% of the time when instituted. That is another hypothesis generating subgroup result, implying further that there may indeed be a role for individualised MAP targets. I like the design of this aspect, with hypertensive therapy as a trial only to be continued in the event of objective improvement.

We have already mentioned that CRT guided resuscitation resulted in less fluid administration and appeared to be even better in the less sick. Worth reiterating that maybe these issues are intertwined. The sepsis police won’t like this, but the sepsis revolution will

Death at 90 days continued to adjust after the primary outcome, with more patients in the CRT group meeting their end over the following 2 months – I am not sure this can be attributed to the interventions within the first 72h. Does it reflect some of the baseline imbalances between groups/

No subgroup analysis favoured lactate clearance to guide resuscitation. All markers of lower severity favoured a CRT guided strategy; lactate <4, SOFA <10 and APACHE2 <25. A lactate guided strategy did not come out of this looking good, no matter how you dressed it up.

And the take home?

Well, in my opinion this trial was clearly underpowered, is difficult to generalise, had issues with recruitment and was at risk of bias in several domains . I am not alone in those views. There is a great interactive tool to examine the sample size issue here.

However, it is hard to read this without acknowledging that the cheaper, easier and more globally transferable assessment of resuscitation effectiveness  outperformed the blood test. And it did so principally by mandating regular review (which is cheap and easy to do) and limiting the fluid administered (which is cheap and easy to do).

What would you want to see before you use this intervention to guide your care? More trials in your population? Statistical significance? Blinded data? I am not sure you need it.  

I for one am going to dust off my use of CRT, standardise my application, pay more attention to it and try to increase the frequency of my clinical reviews. This trial suggests to me another source of harm resulting from the quest for euboxia.

I also think research like this can sometimes remind us about the power of rational repeated clinical assessment and rigorous application of basic principles.

But that’s just me.

What do you think?

Enjoy SMACC if you are going. Some of us have to work……


How you can support St Emlyn’s

Posted by Dan Horner

Dr Daniel Horner BA MBBS MD PgCert MRCP (UK) FRCEM FFICM is an editorial board member on the St Emlyn’s blog and podcast. He is Professor of Emergency Medicine of the Royal College of Emergency Medicine. He is a consultant in Emergency Medicine and Intensive Care at Salford Royal NHS Foundation Trust. He is chair of the national exemplar centre Thrombosis Committee and Regional lead for Injuries and Emergencies on the NIHR Clinical Research Network. He is a Senior clinical lecturer at the University of Manchester and collaborator with the University of Sheffield. You can find him on twitter as @RCEMProf

Thanks so much for following. Viva la #FOAMed

Translate »