No Improvement in Cardiac Arrest Outcomes With Hypothermia
Therapeutic hypothermia, or targeted temperature management, has become a standard component of post-cardiac arrest care. The evidence supporting this practice came from two small-to-medium-sized randomized trials, both published in the New England Journal of Medicine in 2002:
- Among 273 patients with out-of-hospital cardiac arrest due to shockable rhythms (ventricular fibrillation or ventricular tachycardia) at 9 European centers, hypothermia (32-34°) led to improved survival (59% at 6 months vs. 45% with usual care) and better neurologic outcome (55% vs 39% achieving CPC 1-2) (HACA group)
- Hypothermia to 33° led to 49% surviving with a good neurologic outcome (discharge home or to rehab), compared to 26% of normothermic patients, among 77 patients with out-of-hospital cardiac arrest due to ventricular fibrillation in Australia. 49% survived with hypothermia, vs. 32% with usual care (nonsignificant) (Bernard et al)
Guidelines soon arrived, advising the use of hypothermia / targeted temperature management for patients with out-of-hospital cardiac arrest with shockable rhythms. Many centers went further, cooling all victims of cardiac arrest regardless of the "shockability" of the initial rhythm or whether the arrest occurred in or out of the hospital. Skeptics of the new paradigm noted the lack of blinding to treatment allocation in the above studies as a significant limitation (it introduces the potential for biases in treatment between groups). Further, in the larger hypothermia study, a number of patients assigned to usual care developed fever, which is associated with worse outcomes after cardiac arrest.
In the largest randomized trial yet, Niklas Nielsen et al probe further, examining whether hypothermia is effective in patients with and without shockable rhythms, if fever is also prevented as standard care. They publish their results in the online New England Journal of Medicine.
What They Did
Between 2010 and 2013, authors enrolled 950 consecutive victims of out-of-hospital cardiac arrest at 36 centers in Europe and Australia, regardless of the initial rhythm (80% had a shockable initial rhythm, almost always ventricular fibrillation; 12% had asystole and 8% PEA; unwitnessed asystole arrests were excluded). Patients were then randomized to receive targeted temperature management (using any cooling method) to either 33° or 36° C, as soon as possible for 28 hours, followed by rewarming, followed by fever-reduction methods for 72 hours post-arrest.
After 72 hours, a neurologist blinded to initial treatment allocation recommended withdrawal or continued care based on standardized criteria, with withdrawal recommended only for known predictors of a terrible outcome (e.g., refractory status epilepticus; Glasgow motor score 1-2 with bilateral absence of N20 peak on median nerve SSEP).
The primary outcome was mortality of any cause within 6 months of follow-up. Bad neurologic outcome (CPC >2 and Rankin >3, meaning severe disability) or death was a composite secondary outcome, among others.
What They Found
Therapeutic hypothermia to 33° C did not improve outcome in any measurable way. There were no differences between groups in the rate of death (50% with hypothermia, 48% without), or in the composite outcome of poor neurologic outcome or death after 6 months (risk ratios were almost exactly 1). Hypothermia did not lead to any excess of serious adverse events, either.
When the analysis was restricted only to the 80% of subjects with shockable rhythms (a predefined subgroup analysis), there was still no benefit from hypothermia: the relative risk for death among the cooled patients was 1.06.
What It Means
Has all that shivering been for nothing? Has induced hypothermia just taken its place beside intensive glucose control and activated protein C in the sad gallery of fallen hero-therapies in critical care? Unfortunately, the answer is "probably." The simplest and most elegant explanation for the findings is that despite its proven physiologic basis and experimental benefit in animals, therapeutic hypothermia does not improve outcomes meaningfully in people after cardiac arrest, on a population basis. When a much larger, well-conducted randomized trial shows such a conclusive result (statistically speaking) directly refuting smaller trials -- barring methodologic flaws or incomparability -- the result deserves respect. There's a reason they call it "power."
But why the discordance? Conceivably, the previous positive results were created or inflated by the lack of blinding to treatment allocation on the part of the treatment teams -- who, being human, were justifiably excited about the possibility of a new beneficial treatment and their chance to prove its worth. That passion led to unmeasured treatment differences between groups. Or, the optimism infused by assignment to the intervention arm led to a differential rate of care deceleration / withdrawal of care (perhaps as chosen by the patients' families), or differences in care post-hospital discharge.
Alternately, as the current authors suggest with their hypothesis, untreated fevers in the previous trials' control groups could have worsened outcomes; these ill effects of fever were prevented in the present trial. It's hard to evaluate this possibility, because granular data on fever incidence was not included in the published articles.
The other possibility is that the findings in the current study do not reflect truth, and the previous trials (Bernard, HACA) do. One argument for this (as also mentioned by Scott Aberegg in his Medical Evidence Blog) could be the higher "dose" of hypothermia in the prior trials: about 4 - 4.5° average temperature difference between groups, as opposed to 3° in the present trial. But that 1° changing this high-powered, firmly negative trial result seems close to inconceivable.
Treatment teams were not blinded to allocation group here either, and it's possible biases could have also led to care differences between groups. Care was not standardized across all centers, so outlier effects at specific centers could have skewed the overall result from a positive to negative trial. Given the large overall size, number of centers, and total absence of signal, this doesn't seem likely either.
Clinical Takeaway: Temperature should be maintained at 36° C or below after out-of-hospital cardiac arrest. Despite its physiologic rationale and evidence of benefit in prior smaller studies, targeted temperature management below 36° probably does not improve outcomes after out-of-hospital cardiac arrest of any type. Because average human core temperature is 37° C (98.6° F), maintaining temperature continuously at or below 36° C (96.8° F) still would require targeted temperature management (cooling) in almost all patients.
Niklas Nielsen et al, for the TTM investigators. Targeted Temperature Management at 33°C versus 36°C after Cardiac Arrest. N Engl J Med November 17, 2013. DOI: 10.1056/NEJMoa1310519