October has been a stellar month for publication of critical care trials in high impact journals. SPLIT, HEAT and FELLOW have already been FOAMed all over, but that left me with a still tempting choice of EUROTHERM 3235 and EPO-TBI. Given we recruited to the former here in Virchester I thought it would only be polite to give this large undertaking some additional time in the spotlight. After all, the European Society of Intensive Care Medicine did call it “ the most important clinical trial in critical care ever…”, a bold statement if ever I have read one. You’ll have to read the protocol if you want the full quote….
What’s it all about then?
Well, when it comes to neuronal cell death and attenuation of the inflammatory cascade, we in the intensive care unit seem very keen to cool things down. Papers have been published describing the rationale for this therapy, including the EUROTHERM trial protocol if you want to brush up on the pathphysiology. How cool, for how long and in what conditions will complications outweigh benefits all remain key questions. There have been recent lessons from TTM in the case of cardiac arrest but a parallel area where the jury remains out has been the role of therapeutic hypothermia (TH) for neuroprotection following traumatic brain injury. Earlier small studies have led to a position of equipoise. Some centres are using regularly as an adjunct for refractory ICP control, others are more cautious and have asked for further evidence. People often have very strong (dogmatic) opinions either way.
This study aimed to address the equipoise through conduct of a large, multi-centre randomised controlled trial. As always, the full article and accompanying supplementary appendix make good reading and I urge you to have a look at all the data for yourself.
What did they do?
The researchers here designed a study that would randomise patients with recent traumatic brain injury (TBI) and directly measured ICP refractory to tier one strategy control measures, to receive either therapeutic hypothermia for 48 hours (minimum) plus standard care, or standard care only. Tiered neuroprotection is probably worth looking at briefly prior to evaluating this paper if you are unfamiliar. There are online resources available that provide local guidelines to look at as examples. They screened all those with recent TBI admitted to a critical care environment with invasive ICP monitoring and an approach to participate was triggered if the ICP climbed above 20mmHg for 5 minutes refractory to tier 1 management. This varies by hospital, but tier 1 therapy usually includes head elevation, encouragement of venous drainage, intubation, sedation and ventilation to appropriate targets. Regarding timings, they classed ‘recent’ as head injury within 72 hours initially. They reclassified this as within 10 days in January 2012. We will come back to this…..
Randomisation occurred with minimisation through a central control. The trial was open label for therapeutic intervention, but assessor blinded. Centres were directed to cool initially with a 20-30ml/kg bolus of cold saline. They could then maintain hypothermia as they saw fit but directed to a duration of at least 48 hours, with additional time as needed to control ICP. Temperature was optional between 32-35degrees in the intervention arm, titrated to ICP. All patients were followed up for 6 months and the Extended Glasgow outcome scale (EGOS) was used as the primary outcome measure.
That all sounds pretty reasonable?
Indeed. This was a well designed, multi site prospective RCT from the great and the good of the Intensive Care World, reflected by its publication in the NEJM. The researchers powered the trial using an internal pilot, which is also an excellent method of optimising the chances of adequate power – the first hundred or so patients are used as a pilot group to refine the expected difference between groups, which is consequently used within a contemporary power calculation. This is great as these patients are already within the current trial environment. You can’t really get more generalizable to the main trial than that…
Their outcome choice of the Extended Glasgow Outcome Score is also a reasonable one, and assessor blinding should have ensured the avoidance of bias here. However, it is worth noting that they performed an ordinal analysis together with covariate adjustment of the EGOS. This improved the ‘statistical efficiency’ of the outcome data, such that they only required 600 patients when a standard power calculation for a binary outcome (proportion achieving an EGOS of 5-8 vs 1-4 for example) suggested that 1000 patients were required. I’m not entirely sure about the statistical validity of this. However, when you read the full results you can see that power is the least of their worries…
The trialists also engaged a steering committee and a Data Safety and monitoring Committee, with the remit of examining the data regularly to look for episodes of patient harm and evolving likelihood of futility. This is key to ensuring safety in a trial recruiting high risk patients like this one.
The study recruited at over 25 centres; with central allocation it would be impossible to predict allocation as an investigator, so concealment was preserved. The analysis was also intention to treat, which is important to consider in a trial based upon an intervention that can be challenging to deliver in critically ill patients.
So all well and good then?
Well, not really. Although the intention was to recruit 600 patients, the trial was actually stopped after an interim data analysis suggested increasing harm to the hypothermia group.
After 5 years of recruitment and 387 patients, the data committee found a statistically significant increase in the odds of an unfavourable outcome in the group allocated to therapeutic hypothermia (OR 1.53 (1.02 to 2.30)). They also found a statistically significant increase in the odds of death at 6 months (HR 1.45 (1.01 to 2.10)). Not really what they had in mind, but interesting data indeed.
Recruitment ceased at this point, based on the evidence of increased harm with the intervention and the likelihood of achieving futility if the trial were to continue.
Well, that’s a clear result then isn’t it?
I suppose. But there are some key issues to consider in the conduct of this trial. The 2 big ones for me are the place of hypothermia within the tiered escalation strategy for ICP control in TBI and the timing of initiation. The researchers chose to place therapeutic hypothermia at the start of tier 2; I’m not so keen on this, as I think that as a rescue therapy for which equipoise exists, we should be using all other treatments we are comfortable with prior to therapeutic hypothermia. Admittedly, there is also limited evidence for osmotherapy, barbituates and hyperventilation in this clinical context. However, when multiple therapeutic options remain following initiation of the intervention under scutiny, it makes it very difficult to control what happens to the 2 groups after randomisation. Confounding starts to creep in to the analysis. Were the control group more likely to get osmotherapy or barbiturate infusion commenced earlier for example? Could this be the reason for the difference in outcomes?
Secondly, including patients up to 10 days following TBI could be problematic. If you subscribe to the rationale that TH reduces the inflammatory cascade and aims to prevent secondary brain injury, then I would question the validity of introducing it to a patient at day 8 or 9 with established cerebral oedema. Whilst the initial inclusion criteria (<72 hours from injury) seemed very reasonable to me, I struggled to reconcile the pathophysiological rationale with the changed inclusion criteria to recruit patients up to 10 days post injury. The authors performed this alteration to include elderly patients and those with ‘evolving swelling’. However I think what this really does is increase the heterogeneity of their cohort, thus risking a dilution of effect or a false negative outcome. In fact, if you look at appendix S5, you will see that of all patients recruited within 12 hours post injury (which unfortunately was <10% of the overall cohort) mortality was in fact halved in the cohort receiving TH (HR 0.54 (0.07 – 4.03). The confidence intervals are huge here of course and so meaningful results are limited, even though a comparison of the hazard ratios for patients <12h from injury compared to those >12h post suggests a borderline significant difference (p,0.05). But to me it raises the question of whether the design of this study was asking the right question.
I see. So you think there are still questions to be asked here?
I do. I would be particularly keen to look at the effect of TH as a rescue therapy within tier 3 for patients with early (<72 hours) refractory ICP and no further surgical options other than decompression. I also think that questions remain about hypo versus normothermia in this context (Brain TTM if you like), whether patients would do better if hypothermia was not induced by adding 2L of salty fluid to a likely already oedematous brain, and whether a trial including patients based on a cerebral perfusion pressure (MAP – ICP) approach would be superior to one guided by ICP only. One could argue that what these patients need is perfusion of the injured brain, and this is what should be dictating escalation through tiers of therapy. If you recruit based on ICP alone, then you are making regular assumptions about perfusion that may not be accurate.
So, ‘more research needed eh?’ everyone says that about every paper you chicken shit.
Calm down dear. This was a good study in many ways and I think we have a useable clinical result from it. Certainly, for those patients with a head injury over a few days old, who have a steadily increasing ICP refractory to tier 1 therapy, I will not introduce TH at a tier 2 stage. This is not really a change for us, but it is important new evidence in the field of neurotrauma and adds to the puzzle in a great way. Many in the FOAMed world have even stronger feelings about the results.
— Elfyn Thomas (@elfyn_thomas) October 7, 2015
However, will I throw TH out as an option altogether? I’m not sure about that. It does seem to reduce ICP after all, as demonstrated quite eloquently by this paper. I think further work has potential merit and would be interested to see it. For those patients who present acutely with refractory ICP increase and exhausted surgical options, I will be discussing the pro’s and con’s of a trial of TH with my friendly neurosurgeon. If it can avoid a decompressive craniectomy, then I wonder if there are still some occasions where the benefits may potentially outweigh the risks.
That’s all I’m saying, like.
Before you go please don’t forget to…