Prolonged hypothermia improved neurologic outcomes after non-shockable cardiac arrests (HYPERION, CRICS-TRIGGERSEP)
Targeted temperature management (TTM, or therapeutic hypothermia) has become standard therapy after cardiac arrest, especially for ventricular tachycardia and fibrillation--so called shockable rhythms. A new randomized trial has shown that deep, prolonged cooling after cardiac arrest from non-shockable rhythms (PEA and asystole) improved neurologic outcomes. But how robust were the results, and should they change standard practice?
The 2013 Targeted Temperature Management (TTM) Trial showed that among 950 patients, cooling to 33° C after cardiac arrest did not improve survival or neurologic outcomes compared to cooling to 36° C. (That trial's findings have been challenged on the basis that many patients were not cooled quickly.) Since cooling to 36° C is easier than to 33°, it has become a widespread practice in the U.S. after cardiac arrest, especially for shockable rhythms (ventricular tachycardia or fibrillation).
However, only about 20% of patients in the TTM trial had non-shockable rhythms. Follow-up observational studies suggested that some physicians and health systems stopped routinely using hypothermia in patients with non-shockable rhythms, after the TTM trial.
HYPERION Trial (CRICS-TRIGGERSEP)
Authors randomized 584 patients admitted to 25 ICUs in France after PEA or asystolic cardiac arrest to either therapeutic hypothermia to 33° C for ~60 hours, or to normothermia (37° C) for 48 hours. They dubbed their group CRICS-TRIGGERSEP, and their trial HYPERION; it was published in the New England Journal of Medicine.
About 75% of patients’ cardiac arrest were out-of-hospital; most of those were at home. About 80% had asystole as the initial rhythm. More than half of all patients’ cardiac arrest were due to respiratory arrest. Neurologic status, the primary outcome, was determined by a single psychologist conducting structured a phone interview with each patient. “Good neurologic outcome” was defined as cerebral performance status 2 (CPC 2, able to live independently, and possibly to work in a sheltered environment) or CPC 1 (possible mild impairment).
For unclear reasons (perhaps the permuted randomization blocks), ten more patients were randomized to normothermia than to hypothermia; three patients withdrew consent in the hypothermia group, resulting in 13 fewer patients in the final hypothermia group (284 vs. 297).
At 90 days, mortality was statistically equal (81% vs 83% nominally favoring hypothermia), but more patients who received hypothermia had a good neurologic outcome (10.2% vs 5.7%, p=0.04).
On its face, the HYPERION trial shows a number needed to treat of 22 to save one person from disability after a cardiac arrest from a non-shockable rhythm, using prolonged deep cooling. If true, this is a major finding that contradicts what was seen in previous TTM / cardiac arrest trials.
However, the trial will likely spark vigorous debate for multiple reasons:
The fragility index was 1, indicating that a single patient being classified as having had a worse neurologic outcome would render the findings non-significant statistically.
The primary outcome was determined by one psychologist (blinded) based on phone interviews, and by using CPC scores. Modified Rankin scores have been advised instead of CPC, partly because CPC can overestimate neurologic function.
CPC (and all neurologic prognostication tools) also have significant inter-rater variability. In one of the few instances where inter-rater reliability of CPC was tested, it was 70%. If there had been a second psychologist, it is a statistical certainty that disagreements in neurologic assessment would have occurred, potentially changing the outcome.
The control group was maintained at 37° C, rather than 36° C as is the current standard of care in the U.S. Many patients in both groups had fevers. If the control group had been cooled to 36° C, some deleterious fevers may have been prevented and improved outcomes.
The bulk of the observed benefit was driven by in-hospital cardiac arrests; in a pre-specified analysis, the benefit for out-of-hospital arrests was not statistically significant.
As in all cardiac arrest trials, the potential for self-fulfilling prophecy bias was present, because families frequently withdraw care. In HYPERION, most families did so: 304 of the 584 originally randomized patients died from withdrawal of life support, usually within 5 days. By U.S. standards, this is fast, and represents significant cultural and/or health system differences between the two countries.
Eighteen more patients died after life support withdrawal in the normothermia group than in the hypothermia group. This was not because of worse neurologic exams or failure to improve. In 14 of them, “other” with unspecified “comorbidities” was listed as the reason.
It's certainly possible that the larger dose of hypothermia (~60 hours vs. the usual 24, at 33° C instead of 36° C) genuinely improved outcomes after cardiac arrest, and that this should be standard care for patients after PEA and asystole arrests.
However, there is too much potential for chance or bias in HYPERION for us to consider its results as ‘truth’ and use them to drive major practice changes outside of France. It would only take one patient to fall out of the CPC 1-2 category in the hypothermia group (or one to rise into it from the normothermia group) to invalidate the results statistically. Given the expected inter-rater reliability in CPC scores, it’s hard to imagine that a few patients would not have “flipped” had a second psychologist been involved, or the modified Rankin had been used instead.
(Such “flipping” or reshuffling of patient outcomes could have occurred in both directions, and even made the positive findings more robust, although the tendency of the CPC is to over-rate function. The larger point is that substantial uncertainty remains here despite the p value of 0.04.)
The International Liaison Committee on Resuscitation (ILCOR) currently advises patients be cooled to between 32° C and 36° C after cardiac arrest (from any rhythm). The 2015 update of those guidelines offered the 36° C option based on the larger TTM trial's results. ILCOR advises hypothermia for at least 24 hours (weak recommendation, low-quality evidence), with longer periods optional.
Absent a change in those recommendations, or a subsequent confirmatory study, the HYPERION trial may not significantly change the widespread practice in the U.S. of cooling patients to 36° C for 24 hours after cardiac arrest from non-shockable rhythms. That said, the study provides ample reason for individual health systems to cool patients for longer, at lower temperatures, to try to duplicate the benefits seen in the trial.