Reports of therapeutic hypothermia’s death are greatly exaggerated

I expect you’re completely #FOAMed out by the post-publication frenzy stirred up by the TTM Trial.

If not, you’ve come to the right place!

Mike collated the explosion of initial FOAM responses in All in a lather over TTM and there have since been notable additions such as Scott Aberegg’s Chill Out: Homeopathic Hypothermia after Cardiac Arrest… and ICN’s interview with TTM investigators Niklas Nielsen and Anders Aneman one week after TTM. Soon after, Charles Bruen came with his fascinating blogpost/podcast putting the latest studies in historical perspective: Therapeutic hypothermia: The history of general refrigeration.  Meanwhile I’ve been involved in discussions via Google Hangouts, reading what the big brains on the CCM-L listserv have written, been asking questions of one of the authors, and generally trying to get to grips with the strengths, weakness and implications of this study.

So, have we entered the age of ‘homeopathic hypothermia’ as Scott Aberegg suggests tongue-in-cheekily? Or is there still life in therapeutic hypothermia post-arrest?

I’ll start with the bottom line:

If T36C is good enough for Stephen Bernard, it’s good enough for me too (plus he’s my boss)

Bernard post-arrest care

Screenshot from Twitter of Stephen Bernard speaking at the recent ACEM annual scientific meeting (via @HawkmoonHEMS)

Here is the summary of the TTM trial from the newly updated LITFL CCC page on this topic:

  • MCRCT, stratified according to site, no allocation concealment
  • 36 ICUs in Europe and Australia
  • modified intention-to-treat analysis
  • n= 939 (T33C: 473 vs T36C: 466 patients in the primary analysis)
  • Population:
    — inclusion criteria: Age ≥18y, OOHCA of presumed cardiac cause, sustained ROSC for 20 minutes, GCS <8 after sustained ROSC
    — exclusion criteria: . pregnancy, known bleeding diathesis (other than medically induced coagulopathy, e.g. warfarin), suspected or confirmed acute intracranial bleeding or acute stroke, unwitnessed cardiac arrest with initial rhythm asystole, known limitations in therapy and Do Not Resuscitate-order, known disease making 180 days survival unlikely, known pre-arrest Cerebral Performance Category 3 or 4, >4 hours from ROSC to screening, SBP <80 mm Hg in spite of fluid loading/vasopressor and/or inotropic medication/intra aortic balloon pump, temperature on admission <30°C
  • Intervention: TTM at T33C (cooled my various means to target <6hours, maintained T33C for 36h, then rewarmed at 0.25C per hour; fever actively managed until at least 72 hours after cardiac arrest)
  • Comparison: TTM at T36C (otherwise similar treatment to the intervention group)
  • Outcomes:
    — Primary: mortality at 180 days
    — Secondary:  composite of poor neurologic function or death, defined as a Cerebral Performance Category (CPC) of 3 to 5 and a score of 4 to 6 on the modified Rankin scale, at or around 180 days (there were also others such as duration of mechanical ventilation, see below)
  • Results:
    — no difference in mortality: 50% of the T33C (235 of 473 patients) had died, as compared with 48% of the patients in the 36°C group (225 of 466 patients) (hazard ratio with a T33°C, 1.06; 95%CI 0.89-1.28; P=0.51)
    — no difference in neurological outcomes: 54% of the T33C group versus 52% of the 36C group died or had poor neurologic function according to the CPC (RR, 1.02; 95% CI 0.88 to 1.16; P=0.78). Using the modified Rankin scale, the comparable rate was 52% in both groups (RR 1.01; 95% CI 0.89 to 1.14; P=0.87).
    — shorter duration of mechanical ventilation in the T36C group: T33C = 0.83 versus T33C = 0.76 median days receiving mechanical ventilation/days in ICU (P=0.006)
    — serious adverse effects were common and marginally higher (with borderline significance) in the T33C group (93%) compared with the T36C (90%) (RR 1.03; 95% CI 1.00 to 1.08; P=0.09)
    — higher rates of hypokalemia in T33C group (19%) than the T36C group (13%)  P=0.02)
    — no differences found in subgroup analyses: age > 65 years, presence of initial shockable rhythm, time from cardiac arrest to ROSC >25 min, and presence of shock at admission
    — no differences in shivering
    — during the first 7 days of hospitalization, life-sustaining therapy was withdrawn in 247 patients (132 in the 33°C group and 115 in the 36°C group)
  • Commentary and criticisms
    — this study is a methodological masterpiece!
    — unlike Bernard 2002 and HACA 2002, not just VT/VF OOHCA were included (~80% were VF/VT)
    — a useful standardised protocol for neurological prognostication and treatment withdrawal was used
    — survival rates of about 50% may seem high, but remember that this applies to patients who have achieved ROSC for a minimum of 20 minutes}
    — the study was powered to detect a RRR of 20% or an ARR of ~11%, thus the study was not powered to detect a smaller treatment effect (this may be more realistic due to the lower ‘separation effect’ between T33C and T36C compared to the earlier Bernard and HACA studies)
    — less than 50% of T33C patients had reached target at 6 hours, but there was good separation between T33C and T36C groups
    — Baseline balance: higher rates of previous MI and IHD in the T33C group, but no difference in the rates of interventions for these conditions
    — the true patient-orientated outcome that matters is neurologically intact survival, the authors didn’t use this as the primary outcome because mortality is a ‘harder endpoint’ and less subject to bias
    — staff caring for the patients could not be blinded; however the doctors who perform neurological prognostication and data interpretation for the study were
    — TTM differs to the Bernard 2002 and HACA 2002 trials: larger MCRCT with excellent methodology, not limited to VT/VF, control group still received TTM (but at T36C)
  • Bottom line: No difference found between targeted temperature management with a target of T36C compared to T33C

Putting this study in perspective

I’ll say it again, this study is a methodological masterpiece. It differs from the two famous practice-changing but flawed papers published in the NEJM in 2002 (Bernard et al, 2002 and the HACA trial) in that it is much bigger, was not restricted to VT/VF only and had targeted temperature management in both the intervention and comparison group. Of note, the previous studies had no specific protocols for management of patients in the control arm, and included significant numbers of patients that became febrile. Finally, TTM was powered to determine an absolute difference in mortality of 11% between the two groups (a big difference) — there might be a smaller difference that has gone undetected (but there aren’t even any trends or weak signals supporting this) — and it was not powered for the more important patient-centered outcome of neurologically intact survival.

What are the possible explanations for the result of this trial?

TTM might work, at either T33C or T36C, because of avoidance of hyperthermia (leading to decreased metabolic demand and fever-related tissue injury) or simply through a reduction in metabolic demand (e.g. prevention of fever, seizure control, cooling, sedation and neuromuscular blockade).

T33C might conceivably be slightly more effective, but less than this study was powered to detect.  Various physiological plausible mechanisms have been proposed for therapeutic hypothermia, including effects on excitotoxicity, neuroinflammation, apoptosis, free radical production, seizure activity, blood-brain barrier disruption, blood vessel leakage and cerebral thermopooling… As I’m prone to say, the problem with physiological plausibility is that everything is plausible when our understanding of physiology is incomplete.

TTM at any temperature may not confer any benefit, it might simply be associated with the improved overall care resulting from focusing the coordinated efforts of an expert team, with close monitoring and prioritisation of therapies, on a critically ill patient.

So, what do we do now?

As clinicians we need to get together with your colleagues and review our post-arrest protocols. I expect most centers will shift to a target of T36C based on the findings of TTM and the possible benefits of simplifying post-arrest care. One concern I have  is that the patients managed at T36C may be at risk if there is a lapse in temperature control. In other words, these patients may be more likely to develop a fever. Strict adherence to the assigned temperature target should remain a strictly targeted goal of post-arrest care protocols.

Despite this pragmatic solution to the dilemma of what to do now, many uncertainties remain. Even if there truly is no difference as found in this study, there are still numerous unanswered questions:

  • What is the optimum target temperature? (If T33C and T36C are equivalent, what is the shape of the curve on which they lie? T34C might be better, for instance)
  • Does it matter when cooling is started? (prehospital, during CPR, post-ROSC?)
  • What is the optimum duration of targeted temperature management? How long should we actively avoid fever for?
  • What is the optimum rate of rewarming? (less of an issue with TTM at T36C!)
  • Is TTM at any temperature really beneficial over not worrying about the temperature? (previous studies would suggest yes, but how definitive are these?)
  • Should VT/VF be treated differently to non-VT/VF? What about in-hospital cardiac arrests? What about different underlying causes?
  • Should some patients have different target temperatures?
  • What does all this mean in the ECMO era?

The list of questions is daunting to read. Additional RCTs are already underway investigating various aspects of TTM, but I wonder whether traditional RCTs are going to be able to answer many of these questions. Perhaps new trial designs, such as adaptive trials, may provide some of the answers. A researchers work, it would seem, is never done.

Therapeutic hypothermia may look dead at first glance, but what seems like rigor mortis is just a radical facelift. We still have targeted temperature management, but the target has changed to a temperature of 36C.

References and Links


Journal articles

  • Bernard SA et al. Treatment of comatose survivors of out-of-hospital cardiac arrest with induced hypothermia. N Engl J Med 2002;346:557-63. doi: 10.1056/NEJMoa003289
  • The Hypothermia after Cardiac Arrest Study Group. Mild therapeutic hypothermia to improve the neurologic outcome after cardiac arrest. N Engl J Med 2002;346:549-56. doi: 10.1056/NEJMoa012689
  • Nielsen N et al. Targeted temperature management at 33°C versus 36°C after cardiac arrest. N Engl J Med Nov 17 2013 doi: 10.1056/NEJMoa1310519
  • Rittenberger JC, Callaway CW. Temperature Management and Modern Post–Cardiac Arrest Care. N Engl J Med Nov 17 2013 doi: 10.1056/NEJMe1312700
Print Friendly


  1. Matt Astin - @mastinmd says

    Good review of what is out there. I agree hypothermia is not dead. It has undergone a clever renaming process, which politicians here in America are very adept at. TTM should be the standard of care. But, as you put above, we still do not know the exact definition/protocol for TTM. Having read about 150 article on TTM in the past 2.5 years, all I can say is I am looking forward to more.


  2. Lewis says

    Nice one Chris. Your point about the state of nature underlying this sample is true, namely that a smaller benefit than the hypothesis may not be detected. It’s exactly equivalent, though, to say that the hypothesis may be true and simply not reflected in this sample by chance, with a probability of (1-beta) under those assumptions. Another statement which has exactly equal validity is that there may be a larger absolute harm than 2% associated with 33C, which yet remains undetected by the sample. These are all slightly trivial because the current set of data supports only mild harm, and that weakly, in a model trial design which excludes as much bias as possible.

    I think your points about adaptive design and dose response curves are very important because this sort of incremental progress towards understanding of the state of nature over decades makes us forget what the question was!

  3. Ben Margetts says


    I haven’t seen this mentioned in the debate anywhere online and worry that I must have misread my copy of TTM. “After 28 hours, gradual rewarming to 37 degrees C in hourly increments of 0.5 degrees C was commenced in both groups”.

    My previous reading of the literature had lead me to believe that we wanted a rewarming phase rate of 0.15-0.25 degrees per hour. Otherwise the beneficial effects of 33C (the practice formerly known as TH) were lost or at least diluted. This has always been the hardest part of TH to perform properly in my experience.

    Does this make any difference to the interpretation of the results? If not, can I now rewarm at 0.5C/hr given that the outcomes in TTM were at least as good as the treatment groups in HACA/Bernard if I’m going to continue with 33C (which I am, primarily for practical reasons)?

    As an aside, where I work there is no catheter lab (at all) but we still receive OOHCA patients that the local interventional centres don’t want. Maybe we should be trying to fix that part of the “post arrest care bundle” first..?

    • says

      Thanks Ben
      ILCOR guidelines support controlled rewarming at 0.25-0.5C per hour — I think anything in that range is acceptable, but we don’t really know.
      Agree re: cath labpost-arrest -- still a tricky area to negotiate.

  4. says

    Brilliant review of where we are at. Bernard’s talk on the topic at ASM was very interesting, but some off the subtleties come across better in a (well!) written piece such as this. Thanks lad.

  5. says

    Do we need to think about hypothermia at all?

    The targeted temperature management study was methodologically great. Still there are some data that needs clarification.

    1. The number of patients with previous history of IHD was 145 in 33C group compared to 115 in 36C group. Similarly in 33C group there were increased cardiac arrhythmia. Would this have had any impact on the study, as we know that any patient with previous history of IHD has more chances of death?

    2. The number of patients in both groups who had pneumonia was comparable to the number of deaths in both the group. (245 in 33C group and had 235 deaths; in 36 C group the number of pneumonia patients were 214 and no of deaths were 225. (Table 2 in original article, table s12 in supplement)). Data needed but not available- the patients who died- did they all have pneumonia?

    3. Induced hypothermia will lead to reduced immunity and reduced signs of inflammation, hence accurate diagnosis of pneumonia or any other infection not possible in early stages. Hence in these patients targeted hypothermia may be more detrimental than therapeutic. Hence stop thinking about hypothermia at all?

  6. David Denman says

    Whilst the study is excellent, the most interesting thing is the interpretation of the results.

    The study question asked whether cooling to a target of 36 degrees offered a significant survival benefit over 33. The results showed that the new intervention showed no benefit over current practice yet we’re talking about adopting the novel treatment. I would say this is an unprecedented reaction to a piece of clinical research.

    There is an argument that there are other advantages to targeting 36 degrees than survival. This may be the case but this study was not sufficiently powered to prove that there is no difference between the two interventions in the primary outcome. Given the lower than expected event rate it wasn’t sufficiently powered to answer the primary study question.

    This has been well described on Simon Carley’s blog but few people seem to have picked up on it

  7. says

    What if: a.) lower is better; b.) lower leads to more impaired drug clearance; c.) prognostication at 72 hours misclassifies patients with colder patients being misclassified more often as “poor prognosis”? If a, b, and c are true, a difference in favor of colder would be abrogated by the misclassification. Also, if prognostication was done at 72 hours in the study, why does the Twitterpic above show 5 day prognostication?

    • says

      Hi Scott
      Steve Bernard’s slide states 5 day prognostication because in the TTM trial prognostication was 72 hours after the end of the intervention i.e. 108 hours after start of intervention -- or at approximately 5 days (actually 4.5 days in the trial).
      Steve Bernard considers this prognostication time frame to be ‘very conservative’ regarding prognostication the context of therapeutic hypothermia.