Fever is harmful in post-arrest patients. This creates confusion in studies of hypothermia in post-arrest patients: is the benefit of hypothermia due to fever avoidance, due to a benefit from hypothermia itself, or perhaps due to both?
Seminal studies in the early 2000s found mortality benefit from application of hypothermia to 33C after cardiac arrest. However, these studies were flawed due to the lack of adequate temperature management in the control group (where fever was common). This led to decades of confusion over whether hypothermia was truly beneficial, or whether it was just an excessively complex approach to avoiding fever.
The TTM trial in 2013 challenged the benefit of moderate hypothermia, by comparing temperature control at 33C versus 36C.1 No differences were found between the two groups. This suggested that the value of temperature management was simply avoidance of fever. Consequently, numerous international guidelines recommended that post-cardiac arrest temperatures be maintained within a range of 33-36C. Whether 33C or 36C might be superior remains hotly contested.
HYPERION trial: design
This is a multi-center trial of patients with non-shockable cardiac arrest randomized to therapeutic temperature management at 33C or 37C.2 Patient selection criteria are shown here:3
Temperature management strategies are shown below.3 This figure is a bit misleading, because it seems to imply that both groups were exposed to temperature management for the same amount of time (48 hours). In fact, temperature management was performed for ~12 hours longer in the 33C group, due to additional time required for re-warming.
The primary endpoint was good neurologic status after 90 days (defined as Cerebral Performance Category score 1-2, correlating with mild or no disability). This was assessed by a blinded psychiatrist using a semi-structured interview via telephone.
patients and temperatures
Patients are shown below. This was a very sick group of patients, who often presented with asystole (~80%) and frequently received no bystander CPR (~30%). 55% had asphyxia leading to cardiac arrest, a disease mechanism which carries a very poor prognosis. However, 627 patients were excluded from the study because they were deemed to be moribund – so the patients included in this study weren’t truly the sickest of the sick.
Temperatures in both groups are shown below. Some patients in the 37C group did experience fever:
It’s useful to compare this to the temperatures achieved in the TTM trial, shown above. In the TTM trial, tighter temperature control was achieved with less overlap between the two groups. More importantly, in TTM the 36C group didn’t experience fever.
The primary result was positive: more patients in the hypothermia group had a favorable neurologic status at day 90 (10.2% vs. 5.7%, p=0.047 using a Fisher Exact test). Although this result is statistically significant (p <0.05), it barely sneaks in under the wire. Changing a single patient result would result in a p-value above 0.05 (giving the study a fragility index of one). Secondary endpoints were neutral (e.g. mortality).
The fragility of these results is notable – particularly since the study hinged on a single psychologist assessing cerebral performance category via telephone interview. Post-arrest cerebral performance category is known to have substantial inter-rater variability. Using a simple dichotomy of favorable versus unfavorable scores (exactly as was done in HYPERION), 12% of results may be discordant when assessed by different interviewers!4 This makes it likely that the conclusions of the study could have been entirely different if another interviewer performed the assessments.
precision of temperature management & incidence of fever in the 37C group
When examining the temperature curves for this study, two features are notable:
- Some patients in the 37C group did have a fever.
- The precision of temperature control was relatively poor.
These features may be explained by two factors:
- 37C may be too high to allow for a safe margin of error in preventing fever.
- Most patients in the normothermia group were managed using basic external cooling (e.g. cooling blankets), not an adaptive cooling device (table below). Achieving definitive temperature control requires a closed-loop device which senses the patient’s temperature and immediately adapts the applied temperature to prevent any changes (e.g. various commercial devices incorporating a temperature sensor and external cooling pads). Alternatively, simply using cooling blankets cannot achieve precise temperature control, since human response is required and minute adjustments in the applied temperature are impossible.
my four conclusions from this trial
This trial is statistically fragile, so irrefutable conclusions cannot be drawn from it. However, it certainly contains useful information which may help guide our practice. Conclusions from the trial will doubtless vary, but these would be my take-home messages:
#1. Non-shockable cardiac arrest is (still) really bad.
Only 8% of patients in this study survived with good neurologic outcomes, despite outstanding ICU care within the context of a clinical trial. If we consider the 627 patients excluded from the study because they were judged to be moribund, this figure would be even lower (~4%). This may even be a bit better than large observational series obtained outside the context of a RCT.5
#2. Comatose post-arrest patients should be managed with definitive temperature control (using an adaptive-control device) at a target temperature of 33C or 36C (but *not* 37C).
Targeting a temperature of 37C is too high, as this may allow for the occurrence of fever. We’re seeing some gradual creep in the permissiveness of allowed temperatures in post-cardiac arrest patients. 37C seems to be a degree too far.
Furthermore, simply using cooling blankets is inadequate to achieve precise temperature control. Patients should ideally be controlled using an automated closed-loop device which continually measures the patient’s temperature and immediately makes micro-adjustments to the applied temperature. External closed-loop devices with water-adjusted pads are safe and easy to use.
#3. This study provides no information about whether 33C or 36C is preferable.
This study supports the superiority of 33C over 37C. That’s probably true, but also a bit irrelevant. The real debate is between 33C versus 36C. This study very simply provides no information about whether 33C or 36C is superior, because this is not what was studied.
This study is being construed by many as providing direct support for cooling patients to 33C. This is only partially valid. The presence of fever in some patients within the 37C group prevents this study from being able to fully support the use of 33C, because it’s impossible to sort out whether any benefit from 33C reflects the benefit of hypothermia or merely more effective avoidance of fever (the same methodological problem present in the original hypothermia trials from the early 2000s).
Another confounder is that patients in the 33C group were subjected to a longer duration of temperature control. This could affect outcomes, because it’s conceivable that the bioeffective “dose” of TTM relates more to the duration of TTM than the depth of cooling. The duration of physical cooling is especially relevant here, because after discontinuation of cooling many patients developed fevers…
#4. Consider antipyretic therapies to prevent rebound fever after discontinuation of physical cooling.
The temperatures of patients following discontinuation of physical cooling is shown above. Patients in both groups commonly spiked fevers (possibly because the use of antipyretics was discouraged within the study).3 Rebound fever after hypothermia is likely detrimental, potentially undoing much of the benefit of TTM.
Research into TTM has generally focused on the initial temperature (e.g., 33C vs. 36C). Meanwhile, much less attention has been spent on the significance of delayed fevers (and, likewise, exactly how long we should continue tight temperature control).
My preference is to continue scheduled antipyretics (e.g. acetaminophen 1 gram q6hr plus 40 mg prednisone daily) for about a week after cardiac arrest (discussed further here). These simple medical interventions are very effective in avoiding rebound fever. However, this is largely an evidence-free zone, which calls for further study.
- HYPERION trial on the NEJM site here.
- Interpreting the meaning of p values close to 0.05 (PulmCrit; this is particularly relevant for this study)
- Top ten reasons to stop cooling to 33C (PulmCrit)
- Pragmatic comparison 33C vs. 36C (PulmCrit)
- Post-cardiac arrest management (Internet Book of Critical Care)
- Targeted temperature trial changes everything (podcast, EMCrit)
- Post-cardiac arrest care in 2013 with Stephen Bernard: Parts I and II (podcast, EMCrit)
- 1.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;369(23):2197-2206. https://www.ncbi.nlm.nih.gov/pubmed/24237006.
- 2.Lascarrou J-B, Merdji H, Le Gouge A, et al. Targeted Temperature Management for Cardiac Arrest with Nonshockable Rhythm. N Engl J Med. October 2019. doi:10.1056/nejmoa1906661
- 3.Lascarrou J, Meziani F, Le G, et al. Therapeutic hypothermia after nonshockable cardiac arrest: the HYPERION multicenter, randomized, controlled, assessor-blinded, superiority trial. Scand J Trauma Resusc Emerg Med. 2015;23:26. https://www.ncbi.nlm.nih.gov/pubmed/25882712.
- 4.Grossestreuer A, Abella B, Sheak K, et al. Inter-rater reliability of post-arrest cerebral performance category (CPC) scores. Resuscitation. 2016;109:21-24. https://www.ncbi.nlm.nih.gov/pubmed/27650863.
- 5.Andrew E, Nehme Z, Lijovic M, Bernard S, Smith K. Outcomes following out-of-hospital cardiac arrest with an initial cardiac rhythm of asystole or pulseless electrical activity in Victoria, Australia. Resuscitation. 2014;85(11):1633-1639. https://www.ncbi.nlm.nih.gov/pubmed/25110246.