It seems as though hypothermia is the therapy that never quite gets there. The Targeted Temperature Management1,2 trial showed no benefit post ROSC and the RINSE trial 3,4 showed no benefit before ROSC. We’ve also tried hypothermia for trauma and again with little success5. The search for a hypothermia effect in critically ill and injured patients is not simply a fashion of course, th4ere is good laboratory and animal work to suggest that hypothermia should have a beneficial effect. However, it seems that we are struggling to translate those laboratory studies into the real world of clinical practice.
The latest trial of hypothermia looks at patients with status epilepticus and was published in the NEJM in December 20166. The abstract is below, but as always, please go read the full paper on the NEJM website.
What sort of trial is this?
It’s a pragmatic, unblinded randomised controlled trial. That’s an appropriate design (as hypothermia is pretty obvious when you give it you can’t blind patients or clinicians). They did blind the results when analysing them though; this is a form of blinding that prevents those doing the statistical analysis from knowing which patient received treatment or placebo and is a good idea in open label trials like this. Patients were randomsed if they required mechanical ventilation and had ongoing seizure activity for more than 5 minutes, or recurrent seizures without return to baseline. This is an interesting set of inclusion criteria – there is no strict definition of preceding therapy and so some patients had only had 1 drug prior to ventilation. Also there was no mention of EEG inclusion criteria, so the screening process would undoubtedly have missed non convulsive status in paralysed and ventilated patients. However, this is a sub group of the patients we see in the ED at times, who have not responded to initial therapies and need additional care.
What did they do?
Patients in the intervention grop were cooled to 32-34C for 24 hours.Cooling was initiated via cold IV fluid at 4degC and ice packs to the body. This is arguably not the safest or most effective way of inducing hypothermia, as demonstrated later in the paper. Figure 1 suggests it took on average 12 hours post randomisation to hit target temperature.
What were the outcomes?
You could choose many. Length of stay, seizure activity, cost etc. but these are all less important than the one chosen which is the functional outcome of patients. The triallists chose the Glasgow Outcome Scale 7 at 90 days. GOS is a fairly blunt tool but is not an unreasonable one. The scale is outlined below and clearly if it was me I’d like to come out with a score of 5 (which is what the authors chose as the principal outcome measure). A score of 1-3 is clearly a major life event if you started out from a position of no disability at all.
Glasgow Outcome Score.
|1. Death||Severe injury or death without recovery of consciousness|
|2. Persistent vegetative state||Severe damage with prolonged state of unresponsiveness and a lack of higher mental functions|
|3. Severe disability||Severe injury with permanent need for help with daily living|
|4. Moderate disability||No need for assistance in everyday life, employment is possible but may require special equipment.|
|5. Low disability||Light damage with minor neurological and psychological deficits.|
What did they find?
The randomised 270 patients and analysed 268. The first thing to notice really are the baseline characteristics and exceedingly poor outcome of this condition in the cohort. Half of these patients were left with moderate disability or worse at 90 days. We don’t know the baseline GOS, but if we assume it to be reasonable then this is a major burden of disease. There was also a 10% ICU mortality.
Their patients were fairly young with an average age of 57, but only half had a history of epilepsy. Regarding initial treatment, the median time to receive any drug from onset of seizure was 40 minutes. This doesn’t sit well with me at all; a UK paramedic response ambulance time for this condition should be 8 minutes, and there are numerous drug options to administer in a prehospital environment. Also, only just over half the patients received more than one anti-epileptic drug. This sounds rather unusual to me; if I am thinking about an unproven therapy for convulsive status epilepticus (CSE), then I would want to be assured that prior to considering this agent the patient had been managed in accordance with national guidance, to include 2 doses of benzodiazepines, a second line drug agent and then general anaesthesia with thiopentone induction. Reaching for ice cold saline if you have not done all of this already seems premature.
The authors headline result is that there was no significant difference in the proportion of patients with a GOS of 5 at 90 days between the 2 groups. 49% in the treatment group vs. 43% in the standard group. There were slightly more adverse events and more difficulties in managing the hypothermia group (as you would expect).
There were some interesting secondary outcomes and subgroup analyses. For instance, all patients had EEG monitoring commenced 2 hours after randomisation and continued for 48 hours. If you look at EEG progression to Status epilepticus, it was halved by therapeutic hypothermia with a statistically significant odds ratio at 0.4 (0.2 to 0.8). Likewise, the chances of refractory or super-refractory status were reduced in the TH group, although not significantly.
The other interesting subgroup was when they divided age – if you assume that elderly patients are likely to have a poorer outcome and look at the under 65’s predominately (of which there were 184), then the forest plot in figure 2 shows a fairly impressive odds ratio for a good outcome which is on the cusp of statistical significance.
Well yes. Although this study is large in the field it’s still pretty small. There was a non statistically significant treatment effect of 7% which if true could be clinically important and yet a study of this size would have to be a lot larger to detect a difference of this level. In fact the authors powered this study to detect an absolute risk difference of 20%. That’s an ambition to detect a number needed to treat of 5. Really? There are very few therapies that can deliver such a huge treatment effect and so I am concerned that this study is significantly underpowered. In addition, there was no accounting for dropout/crossover in the sample size calculation and there is no reference for the suggested improvement of 20% with TH.
In addition, the lack of standardised treatment protocols and /or EEG screening prior to randomisation, long time to any initial treatments and induction of hypothermia with cold IV fluids make me concerned about the external validity of the trial. All patients receiving TH were also paralysed; presumably to prevent issues with shivering, but this introduces a potentially harmful treatment to the intervention arm only. We certainly do things very different locally. We should also mention that the control group received targeted normothermia – this is also likely to make any outcome differences harder to find, but presumably was chosen as the current standard of care.
Now, at St.Emlyn’s we are very wary of anyone who reports alternative outcome data to that agreed a priori. However, it’s sometimes interesting to look at other potential ways to deal with the data and in the supplement you can look at the distribution of GOS scores between those in the hypothermia and normal care groups. Interestingly if the authors had chosen a GOS of 4 or 5 then the difference is greater (9% difference), but again not enough to even approach statistical significance.
I am thus a little wary that this trial is underpowered to detect a meaningful clinically important outcome. Perhaps we cannot abandon hypothermia just yet.
The bottom line.
This trial failed to find a difference in 90 day outcome when hypothermia was used in patients with status epilepticus. We can’t routinely recommend it as a result, but we are concerned that there were simply too few people in this study to answer a remarkably ambitious sample size estimate8. Also, it’s a bit challenging to be clear about what this trial was testing – if It was the role of hypothermia in super-refractory CSE, then we think it needed a protocolised management plan prior to initiation. If it was hypothermia in all fitting ICU patients who had been ventilated, then we wonder if EEG screening criteria should have been included. As such you are left with an unsurprising result of no significant difference found in a mixed cohort in an underpowered study. The story of critical care trials at the moment if you believe Jean-Louis Vincent……
Dan and Simon
NB. Great minds think alike so please check out The Clinical Bottom Line review here.9
EMNerd has also published a review on Hybernatus here10
Before you go please don’t forget to…
- Subscribe to the blog (look top right for the link)
- Subscribe to our PODCAST on iTunes
- Follow us on twitter @stemlyns
- See our best pics and photos on Instagram
- PLEASE Like us on Facebook
- Find out more about the St.Emlyn’s team
5 thoughts on “Hypothermia in status epilepticus. St.Emlyn’s”
Pingback: Hypothermia in status epilepticus. St.Emlyn’s – Global Intensive Care
Pingback: HYBERNATUS – The Bottom Line
Interesting post on an under-evidenced area of practice, thanks.
I’m curious as to why you advocate sodium thiopental induction? There doesn’t seem to be much evidence in favour (e.g. the very limited Cochrane review at: http://www.cochrane.org/CD009202/EPILEPSY_propofol-versus-thiopental-sodium-for-the-treatment-of-refractory-status-epilepticus-rse).
I think I’d probably be keener to use Propofol out of sheer familiarity.
A good point.
I don’t know of any clinical evidence of benefit for one induction agent over the other in these patients. As such, most national guidelines would advocate general anaesthesia, once you reached that stage, with Thiopentone, Propofol or Midazolam. An example guideline can be found here http://www.epilepsycurrents.org/doi/pdf/10.5698/1535-7597-16.1.48?code=amep-site
In the UK, Thiopentone has a fairly established anecdotal role in neuroprotection, mostly based on animal data regarding cerebral blood flow and CMRO2. As such we use it a lot and have familiarity/experience.
Also as a neurointensivist, I use thiopentone episodically to burst suppression in tier 3 management of raised ICP, and as a temporising measure for ICP spikes on transfer. It seems to be one of the few things that will lower directly monitored ICP in patients already established on high dose propofol and with robust tier 1/2 measures already in place. As such I am relatively swayed by the anecdotal evidence and it is my first line agent for neuro emergencies with cardiovascular stability.
I use a lot of propofol for sedation, but tend to avoid it when I am intubating a critically ill patient. I always think there are better options for these patients.
Having said that, sometimes the safest drug is the one you are most familiar with. So I appreciate there are several ways to skin a cat.
Thanks for reading and taking the time to comment.
Pingback: SGEM#199: Therapeutic Hypothermia – What is it Good For? | The Skeptics Guide to Emergency Medicine