A Crack in the Ice? An In-Depth Breakdown of the TTM Trial

PICT0014

Like many other Emergency Medicine residencies, we took the time in our last Journal Club to break down the Targeted Temperature Management Article.  There is tons out there in the #FOAMed space about this trial.  And, one of our 4th year residents, Dr. Trent Wray, took some extra time to break down the article in gory detail and put it into the context of the previously published literature.

Nielsen N, Wetterslev J, Cronberg T, et al. Targeted temperature management at 33°C versus 36°C after cardiac arrest. N Engl J Med 2013

Background

• Since the two landmark studies in 2002 were published (Bernard et al, HACA study group), Therapeutic hypothermia (TH) has become the standard of care after ventricular tachycardia/fibrillation (VT/VF) out-of-hospital cardiac arrests.  A few studies (Hachimi-Idrissi) and a Cochrane review (2012) have reinforced improved outcomes since 2002.  The AHA has given it a class I recommendation (one of the very few class I recommendations it gives), and its use has been extended to other rhythms (PEA, asystole) as well as in-hospital cardiac arrests (albeit with little data to support it).

• Though other studies somewhat support TH, we mainly use the HACA trial and the Bernard trial as the evidence for what we do, except we apply it to non-shockable rhythms as well, citing that the trials are probably way too small (so few survive with non-shockable rhythms) to show a difference when there might be one.

• For the last few years, some have pushed for more study, correctly citing the lack of temperature control in the control group of the HACA and Bernard trials as a potential issue.  Since post-arrest patients have a robust inflammatory response and propensity to develop fevers, there is both a high incidence and, possibly, an impact of post-arrest fever on neurologic outcome.
  • Post-arrest fever has been associated with worse outcomes (a couple of articles published in recently in Resuscitation support this as well).
  • Nielsen argues that, since the main studies we use to support the widespread use of HACA did not control for this potentially detrimental variable in the control group, this should be retrialed with this included.  Thus, this trial sought to ask: in patients with out of hospital cardiac arrest, does cooling them to 33⁰C vs 36⁰C make a difference?
    • Is the “hypothermia” aspect of the “therapeutic hypothermia” making the difference, or is it just that we were finally treating temperature as a variable that should be controlled?

Methods:

• Funding:
  • NO COMMERCIAL FUNDING (can be important)
  • Timing:
    • 2009-2012
  • Setting: 36 centers across Europe and a few in Australia.
    • Most contributed < 25 total patients (one contributed 2).
  • Patients:
    • 950
    • Inclusions:
      • Age > 18
      • Out of hospital cardiac arrest – “presumed cardiac cause” with ROSC for > 20 minutes
      • GCS < 8 (correlates to “not following commands” in this setting)
      • Could be randomized to cooling in 240 minutes (4 hrs) after ROSC
    • Exclusions:
      • Unwitnessed + asystole
      • Body temp < 30◦C
      • Trauma as the cause
      • Pregnant
      • Known inherent coagulopathy (iatrogenic coagulopathy was included)
      • Known terminal illness
  • Intervention:
    • 1:1 randomization – 473 in 33⁰C group, 466 in the 36⁰C group
    • Temp measured by bladder catheter or esophageal probe (if anuric)
    • Both groups received the intervention – either they were cooled to 33⁰C or maintained/cooled to 36⁰C
    • Interesting side note: the incidence of shivering was about the same in both groups (30% in the  33⁰C group vs 34% in the 36⁰C group).
    • After 28 hours from ROSC (24 hours at the target temp), they were allowed to rewarm at 0.5⁰C/hr until they reached 37⁰C.
    • After 72 hours, there was an “intention” for them not to get a fever, though it was left to the sites how they controlled temperature after that.
    • Continued maximum care for 108 hours, and they were banned from withdrawing care for neurological reasons prior to this, UNLESS (mostly ethical reasons):
      • They found out they had a terminal condition (ex widely disseminated cancer, etc)
      • Refractory shock with multi-organ failure despite maximum therapy (i.e. physiologic futility)
      • Brain Death + Early Myoclonus + no N20 peak on SSEP

*Note: 16% in the 33⁰C group and 13% of the 36⁰C group died prior to neuroprognostication, with a few more in the 33⁰C group having a “hemodynamic cause”

*Note: 44% in the 33⁰C group and 52% of the 36⁰C regained consciousness prior to 108 hrs

• Neuroprognostication occurred at 108 hours (72 hrs after rewarming) – they could withdraw care IF:
  • Brain death due to herniation
  • Severe myoclonic status and no N20 peak on SSEP
  • GCS motor score of 1-2, no Median Nerve N20 peak on Median Nerve SSEP, and status epilepticus had been treated

If they met all of the criteria, but they didn’t wake up (i.e. GCS motor score of 1-2, but they HAVE an N20 peak on median nerve SSEP), they reexamined daily and tried to figure out whether or not their sedation was causing it.  Once they decided their sedation wasn’t causing their continued low GCS motor score, they could withdraw care.

Based on the protocol (again, refer to the appendix), recommended to withdraw care in 48% of the 33⁰C group and 42% of the 36⁰C group.

*Note – it is uncertain whether or not 108 hours is the best time, but this is one of the best aspects of the study.  It stems from the large number of patients that die in the setting of withdrawal of care after cardiac arrest (including a large number in the HACA trial).  I will point you to the appendix of the trial for anything more than the basics above.

• Follow-up:
  • At 180 days (6 months) – most were face to face, some were called (achieved 98.6% 180  day f/u)
• Stats:
  • Wilcoxon signed-ranked test, logistic regression, Cox analyses, Odds ratios à relative risk, p-values at 0.05 = statistical significance.
  • Note: goal was to detect an 80% chance of a 20% difference.
    • The HACA trial did NOT find this much of a difference
    • Would have needed thousands of patients to detect a 5 or 10% difference (ex. tPA for stroke is a 6% difference at best, CRASH 2 showed a 1.5% mortality benefit with TXA).
    • Statistical significance aside, we can say that both groups (both of which had interventions in the control of their temperature) essentially did the same, which is, purely per the results, much different than the HACA trial.
• Outcome Measures:
  • Primary: All cause mortality at 180 days (6 months)
  • Secondary:
    • All-cause mortality + Cerebral Performance Category (CPC) 3-5
    • All-cause mortality + Modified Rankin Scale (MRS) 4-6
    • CPC at 180 days
    • MRS at 180 days
    • Adverse events (up to 7 days)
  • Tertiary:
    • “Best” neurologic outcome
    • Neurologic function – they did an MMSE on these folks
    • “Complete Neurologic Recovery” – they just asked them
    • CPC at hospital discharge
• Results:
  • The major results were:
    • Mortality at 180 days – 49% in 33⁰C group and 47% in 36⁰C group
    • CPC 1 or 2: 46.4% in 33⁰C group vs 47.8% in 36⁰C group
    • CPC 3-4: 5% in 33⁰C group vs 5% in 36⁰C group
    • Complications: none statistically significant (see below).  Major ones were fairly low in general.
    • Author’s Conclusions:
      • Their own cited limitations:
• ICU staff wasn’t blinded
  • Those who performed all follow-up assessments were blinded
  • Hard to be, as they were primarily taking care of the patients
• No data on what kind of sedation was used:
  • May have made a small difference in neuroprognostication.
  • Unlikely, but this is a potential confounder when it comes to the protocol on withdrawal of care
• 48% mortality in the “control” group, and better CPC of the control group:
  • This does indicate far better overall intra-and-post arrest care (for example, the one-minute downtime prior to CPR) and MIGHT indicate the treatment group in the HACA trial may have received better overall cardiac arrest care than the control.
  • HACA was 55% mortality, 39% CPC 1 or 2 in the control group
  • The study populations may have been different than the HACA trial

My Critique

     This was very well-done in nearly all aspects.  They tried to make it as real-life as possible, and even included a great protocol for neuroprognostication, which, given that the vast majority of deaths post-arrest are due to withdrawal of care, I think is one of the best aspects of this study.  If you read the appendix, it’s very thorough.

     The 33⁰C group may have been a touch sicker, and more had prior CAD.  In the HACA trial, more in the control group had CAD and more received bystander CPR.  And, I'm not sure of the significance of any of this.  The neuroprognostication is interesting.  As I understand it, they performed neuroprognostication at 108 hours after the arrest (72 hours after rewarming, which is consistent with our goals.  Despite the use of a recently proposed (Cronberg, 2013) protocol in both groups, 72 hours after rewarming may favor the control group, as the sedatives in hypothermia may last longer.  I do think this is unlikely to make a difference, given their criteria (and note the study by Grossestreeuer below), but there’s a small chance it could have.

     It is interesting that they didn’t make CPC 1 or 2 as the primary outcome (the authors do explain why in the appendix – in the world of post-arrest management, mortality is a harder outcome to statistically achieve).  Still, they included enough data to compare it with the HACA trial, and really the main difference is that their control group, which also had their temp controlled, was much better than the control group of the HACA trial.  To me, that’s less of a limitation than it is a support of the study’s thought process (the control of fever was one of the limiting factors of the HACA trial).

A Note the “Power” Calculation:

• In their protocol, they state they wanted to get to the population needed to have an 80% chance of finding a 20% difference between the 33⁰C group and the 36⁰C group (10% relative risk reduction).  This wasn’t TOO far off from what the HACA trial found, but it’s still pretty bold.  Still, there wasn’t even much of a trend toward a benefit, so not real sure it matters.
•  That said, to find a 5% difference (which would be in-line with the NINDS tPA trial for stroke), or even a 10% difference, they would have needed thousands of patients.

Thus, to satisfy the stats folks, THIS IS NOT A STUDY THAT SHOWS “NO DIFFERENCE”!!!!

• Just because the study didn’t detect a 20% difference, doesn’t mean there isn’t a difference at all.
• For all we know, there is a chance that cooling to 33⁰C provides harm too (slight trend that way).
• It’s going to be very tough to get to a sample size to know this
• Statistics aside, both treatments had essentially the same effect on outcome, which is what is important.
• The study COULD NOT SHOW A BENEFIT OF COOLING TO 33⁰C vs 36⁰C in their study population.

Summary of Arguments for and Against

IS THERE A BENEFIT TO COOLING TO 33⁰C??

On one-hand – NO: Not ready to change practice, as many will still benefit from lower temps – we should still cool to 33⁰C??

Hard to call a “fever” in the HACA trial – the mean was 37.8⁰C

Though many patients likely became hyperthermic, the mean didn’t have a true “fever”.  It’s hard to say whether the author’s proposition that control of fever could provide the same benefit as cooling to 33⁰C based on this study.

The study may have been underpowered to detect a difference – compare it to CRASH 2, which had thousands in it to detect a 1.5% difference in mortality (and we’re all ready to jump on the bandwagon of TXA).

8% of eligible patients in the HACA trial were included, while over 2/3 of eligible patients in the TTM trial were included.  This may have been due to better trial design (i.e. faster randomization to meet the 4 hour window) or it could be that some folks do benefit, and the HACA trial just found them.

Not sure the “no benefit” applies this to our patients (folks in Europe who get CPR within one minute of collapsing? That is NOT the norm here):

• Note: the patients in the HACA trial were different as well, and we used this with no problem before.
• The mean time to CPR was 1 minute in this trial.  Most of our patients don’t get that.  In the HACA trial, about half even received bystander CPR at all (though, the total downtime before ROSC was the same in the 2 trials).  Does this mean those with the longer downtime prior to CPR benefit more?
• This will be difficult to study, as so few adults with long downtimes survive anyway.

Should we just keep using it, then, as most of our patients have longer downtimes?? The mean age of the patients in the newer TTM trial was older than the HACA trial (though about the same in the Bernard trial).   What about younger patients in general?  Will they benefit more?

• Neither trial used endovascular cooling all that much (25% in the TTM trial, all surface in the HACA trial).  Both trials had some swings in temperature (in both groups in each trial).  We know swings in temperature can be bad for a brain-injured patient, and cooling to 33⁰C likely results in more wide swings of temperature (to the point it might negate the protective effect of hypothermia).  Maybe cooling in a more controlled fashion with an endovascular device (i.e. we waited until they got to the ICU and cooled them with iced saline, then an endovascular device only) would a lower temp be more helpful then?

• There have been times, when we’ve discontinued a therapy, only to have it restarted again due to a new study.  The best example of this is the PRONE-SUPINE study for proning in ARDS.  These resulted in a decrease in the use of proning (particularly by less experienced centers), but a new study earlier this year showed a difference when it was applied earlier and by experienced centers. We shouldn’t make the same mistake by decreasing the use of hypothermia based on this study.

Though it could not show a benefit, the trial didn’t show harm either, and there are many subgroups of patients that may still benefit (especially those who do not get immediate CPR, as most of our patients don’t).  The goal should still be 33⁰C  for all patients.

On the other-hand – YES, THIS STUDY IS NOW THE BEST AVAILABLE EVIDENCE.  WE SHOULD CHANGE THE TARGET TEMP TO 36⁰CFOR ALL POST-ARREST PATIENTS:

• This study is extremely well-done.  It applies hypothermia how we have been applying it in real-life (to all rhythms), but adds a control group where the control-group temperature is just slightly higher.  It is now the best available evidence.

• Remember, the 1 minute downtime to CPR is NOT the total downtime prior to ROSC, which was the same in both groups.

• Though this study DOES NOT SHOW “NO DIFFERENCE”, we should not apply a therapy that, according to the new best available evidence, ISN’T SHOWN TO BENEFIT.  We usually stop using therapies routinely that DO NOT SHOW A BENEFIT.  The best example of this is steroids in sepsis.

• Even the HACA trial, which showed a benefit, did not reach target temperature until a mean of 8 hours, and actually did not reach target temperature in many of their patients – it’s likely NOT the cooling, but prevention of hyperthermia.

• Even though none of the complications reached statistical significance, there is a trend toward increased complications (pneumonia, bleeding, arrhythmias) with lower temperatures.  Also, it might use more resources.  Why shoot for lower temperatures if it doesn’t benefit?

• We based nearly all of our existing cooling practices (other than extending it to non-shockable rhythms and in-hospital arrests…mostly because it’s tough to get enough patients to truly study) on the HACA trial prior to this study. The control group in this study had a neuro-favorable outcome of 39%.  Most striking to me is the improvement in the control group.  It went from 39% CPC 1 or 2 the HACA trial to 48% in the TTM trial.

• It may be more about good post-arrest care than anything.  We don’t necessarily need lower temperatures for that.

• CONTROL is part of good post-arrest care.  Prevention of wide swings in target temp in what is essentially a brain-injured patient.

• Even if TH provides a benefit, it’s likely NOWHERE NEAR the benefit provided by just keeping temperature in strict, tight control along with good post-arrest care.

• Currently, we have a study that is three times larger than what we based nearly all of our practices on before, encompasses the best, most modern critical care possible for post-arrest patients , and it shows NO BENEFIT to cooling patients to 33⁰C  vs just keeping their temperature STRICTLY CONTROLLED at 36⁰C.  This is the best available evidence, and it shows no benefit.  There’s no reason to keep cooling to 33⁰C.  We should focus on good overall intra-arrest and post-arrest care, and just control the temp at 36⁰C.

What it Means for Us…for Now

• We’re still trying to minimize brain-injury – good, expedient intra-arrest and controlled, targeted post-arrest care does that.

• We still believe a target temperature of 33⁰C will provide the best overall outcome in our setting and for our patient population.

• We need to do this in a CONTROLLED fashion.

• Prehospital cooling is likely not a good idea (based on the JAMA study published just before this trial), and may CAUSE swings in temperature we know are harmful in brian-injured patients.  DO NOT INDUCE HYPOTHERMIA IN AN UNCONTROLLED ENVIRONMENT.

• No rush to get them to target temperature (it’s not the speed, it’s the control).  The goal is within 4 hours, and earlier has not been shown to be any better.  Just make sure you do it, and do NOT allow swings in temperature.
• Place a temperature-sensing foley right away (or esophageal probe if they’re not making urine).
• If they’re > 33⁰C, go ahead and place ice packs, you may give 30 mL/kg cold saline as well – do NOT allow them to accidentally rewarm.
• Avoid rapid, wide swings in temperature – we know these are bad for brain-injured patients.
• Avoid fever – this is also likely bad for brain-injured patients
• Keep tight control on every variable you know is important.
• Avoid wide swings in temperature
• Minimize wide swings in MAP (try and keep it 70-110)
• Minimize swings in PaCO2 and pH, do not hyperventilate
• Minimize hypoxia and hyperoxia
• Manage hyper- and hypoglycemia
• Aggressively treat seizures.
• Sedate and provide paralysis
• Push toward normalization of lactate

Consult Neurocritical care:

• For now, based on the above debate, there’s no reason to change our current practice.
• The goal is still 33⁰C for all cardiac arrests that don’t meet exclusion criteria.
• NCC will likely still place an endovascular device.  There is no evidence for this, but, based on principle, we need to minimize swings in their temperature.  At our site, we do this best with a bladder catheter and an endovascular cooling device.
• Even if we don’t cool to 33⁰C (ex. they’re not a candidate for whatever reason), STILL KEEP THEM AT 36⁰C regardless of whether or not they’re in the cooling protocol.
• If, for any reason, the 33⁰C protocol is stopped (as it was in 14 of the patients in the TTM trial), the plan will be to continue the protocol at 36⁰C, as it is associated with less complications and we KNOW that control of the fever is likely the most important variable.
• For all patients, the protocol should be continued for 72 hours with NO allowance for fever.

References:

1. Arrich, J, Holzer, M et al (2012).  “Hypothermia for Neuroprotection in Adults after Cardiopulmonary Resuscitation.” Cochrane Database Syst Rev 2012;9:CD004128
2. Bernard, SA, Gray, TW et al (2002).  “Treatment of Comatose Survivors of Out-of-Hospital Cardiac Arrest with Induced Hypothermia.”  NEJM 2002 346(8) 557-563
3. Bro-Jeppesen, J, Hassager, C et al (2013).  “Post-Hypothermia Fever is Associated with Increased Mortality after Out of Hospital Cardiac Arrest”.  Resuscitation 2013(Dec);84(12)1734-1740
4. Caufield, F, Rachabattula, S et al (2011).  “A Comparison of Cooling Techniques to Treat Cardiac Arrest Patients with Hypothermia.”  Stroke Res Treat 2011;2011:6090506
5. Cronberg, T, Horn, J et al (2013).  “A Structured Approach to neurologic Prognostication in Clinical Cardiac Arrest Trials.”  Scandinavian Journal of Trauma, Resuscitation, and Emergency Medicine 2013;21(45)1751-7
6. Diao, M, Fenglou, H et al (2013).  “Prehospital Therapeutic Hypothermia after Cardiac Arrest: a Systematic Review and Meta-Analysis of Randomized Controlled Trials.”  Resuscitation 84(2013):1021-1028
7. Flint, AC, Hemphill, JC et al (2007).  “Therapeutic Hypothermia After Cardiac Arrest: Performance Characteristics and Safety of Surface Cooling With or Without Endovascular Cooling.”  Neurocrit Care 2007;7(2)109-118
8. Gillies, MA, Pratt, R et al (2010).  “Therapeutic Hypothermia after Cardiac Arrest: a Retrospective Comparison of Surface and Endovascular Cooling Techniques.”  Resuscitation 2010 Sep;81(9):117-112
9. Grossestreuer, AV, Abella, BS et al (2013).  “Time to Awakening and Neurologic Outcome in Therapeutic Hypothermia-Treated Cardiac Arrest Patients.”  Resuscitation 2013 Dec, 84(12):1741-1746
10. Kim, F, Nichol, G et al (2013).  “Effect of Prehospital Induction of Mild Hypothermia on Survival and Neurological Status Among Adults with Cardiac Arrest.”  JAMA 2013 doiL10.1001/ama.2013.282173
11. Legatt, Alan (2012).  “General Principles of Somatosensory Evoked Potentials.” Medscape emedicine.medscape.com/article/1139906-overview (accessed 12/4/2013)
12. Hachimi-Idrissi, S, Corne, L et al (2001).  “Mild Hypothermia Induced by a Helmet Device: a Clinical Feasibility Study.”  Resuscitation 2001 Dec;51(3):275-281
13. Leary, M, Grossestreuer, A et al (2013).  “Pyrexia and Neurologic Outcomes after Therapeutic Hypothermia for Cardiac Arrest.”  Resuscitation 2013(Aug);84:1056-1061
14. Nielsen, N, Wetterslev, J et al (2013).  “Targeted Temperature Management at 33⁰C vs 36⁰C after Cardiac Arrest.”  NEJM 213:369:2197-2206
15. Peberdy, MA, Callaway, CW et al (2010).  “Post-Cardiac Arrest Care: 2010 American Heart Association Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care.” Circulation 2010;122;Suppl3:S768-S786
16. Pfeifer, R, Weitzel S et al (2013).  “Investigation of the Inter-Observer Variability Effect on the Prognostic Value of Somatosensory Evoked Potentials of the Median Nerve (SSEP) in Cardiac Arrest Survivors Using an SSEP Classification,”  Resuscitation 84(2013) 1375-1381
17. Rittenberger, JC, Raina, K et al (2011).  “Association between Cerebral Performance Category, Modified Rankin Scale, and Discharge Disposition after Cardiac Arrest.”  Resuscitation 82(2011):1036-1040
18. The Hypothermia After Cardiac Arrest Study Group (2002).  “Mild Therapeutic Hypothermia to Improve the neurologic Outcome after Cardiac Arrest: NEJM 2002 346(8)
19. Tomte, O, Draegni, T, et al (2011).  “A Comparison of Intravascular and Surface Cooling Techniques in Comatose Cardiac Arrest Survivors.”  Crit Care med 2011 Mar;39(3):443-449
20. Wijdicks EF, Hijdra A, Young GB, Bassetti CL, Wiebe S,. Practice parameter: prediction of outcome in comatose survivors after cardiopulmonary resuscitation (an evidence-based review): report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology. Jul 25 2006;67(2):203-10
21.  Zeiner, A, Holzer, M et al (2001).  “Hyperthermia after Cardiac Arrest is Associated with an Unfavorable neurologic Outcome.”  Arch Int med 2001;161:2007-2012.