We conducted this investigator-initiated, open-label, blinded-outcome-assessor, pragmatic, multicenter, randomized, controlled trial in 25 intensive care units (ICUs) in France (11 in university hospitals and 14 in community hospitals). The trial rationale and design have been described previously. The research
protocol (available with the full text of this article at NEJM.org) was approved by the appropriate ethics committees and French data-protection authorities.
The trial was supported by independent research grants from the French Ministry of Health, the nonprofit health care institution Centre Hospitalier Départemental Vendée, and the Laerdal Foundation. None of the trial funders had any role in the trial design, the collection or analysis of the data, or the writing of the manuscript. The authors vouch for the completeness and accuracy of the data and for the fidelity of the trial to the protocol.
Candidates for trial enrollment were 18 years of age or older and had been resuscitated from out-of-hospital or in-hospital cardiac arrest with a nonshockable rhythm due to any cause. Eligible patients had coma (score ≤8 on the Glasgow Coma Scale [GCS]; scores range from 3 to 15, with lower scores indicating poorer function) at ICU admission. In patients who had been sedated before ICU admission, the GCS score that had been determined by the emergency physician just before sedation was used. Exclusion criteria were a no-flow time (from collapse to initiation of cardiopulmonary resuscitation [CPR]) of more than 10 minutes; a low-flow time (from initiation of CPR to return of spontaneous circulation) of more than 60 minutes; major hemodynamic instability (continuous epinephrine or norepinephrine infusion >1 μg per kilogram of body weight per minute); time from cardiac arrest to screening of more than 300 minutes; moribund condition; Child–Pugh class C cirrhosis of the liver (severe hepatic dysfunction); pregnancy or breast-feeding; status of being under guardianship; status of being an inmate at a correctional facility; previous inclusion in another randomized, controlled trial involving patients with cardiac arrest in which the neurologic outcome at 90 days was assessed as the primary end point; lack of health insurance; and decision by the next of kin for the patient not to participate.
According to French law, because the strategies used in both groups were considered to be components of standard care, informed consent for trial participation was not required. However, French data-protection authorities require that patients be given the opportunity to decline that their data be used. Therefore, since the patients had coma, it was required that the closest available relatives receive specific information about trial enrollment. Patients with no available relative were included in the trial, were informed as soon as they regained competence, and were asked whether they wanted to remain in the trial; if the answer was negative, they were excluded from the analysis.
RANDOMIZATION AND BLINDING
At each center, eligible patients were randomly assigned in a 1:1 ratio to either moderate therapeutic hypothermia (33°C) or targeted normothermia (37°C). Randomization was conducted with the use of a Web-based system that was accessible 24 hours per day. The randomization sequence was generated by the statistician (who was not involved in the recruitment of patients), with permuted blocks of varying sizes and with stratification according to center and cause of cardiac arrest (presumed cardiac vs. presumed noncardiac).
It was not feasible for the staff who were providing patient care to be unaware of the group assignments. However, the psychologist who assessed the trial outcomes in all patients was unaware of the group assignments.
TARGETED TEMPERATURE MANAGEMENT
The trial protocol involved the standardization of several variables including sedation, neuromuscular blockade,
and the management of expected adverse events. Details are provided in the
Supplementary Appendix, available at NEJM.org.
In patients who had been assigned to the hypothermia group, hypothermia at 33°C (with a window of ±0.5°C) was induced and then maintained for 24 hours. Each center followed its standard protocol (active internal cooling with a specific device, active external cooling with a specific device, or active external cooling without a specific device). Slow rewarming was then performed at a rate of 0.25 to 0.50°C per hour, to 36.5 to 37.5°C, which was maintained for 24 hours. Sedation was provided according to the standard protocol in each center, with dosage adjustment to obtain a Richmond Agitation–Sedation Scale score of −5 (on a scale from −5 [unresponsive] to +4 [combative]).
During rewarming, sedation was tapered when the body temperature rose above 36°C.
In patients who had been assigned to the normothermia group, body temperature was maintained at 36.5 to 37.5°C for 48 hours according to the standard protocol in each ICU. Sedation was given routinely only during the first 12 hours after randomization, in accordance with 2010 ILCOR guidelines.
ASSESSMENT OF NEUROLOGIC PROGNOSIS AND LIFE-SUPPORT WITHDRAWAL
Decisions regarding limitation of treatment followed current guidelines.
A multimodal assessment of neurologic prognosis was performed, with the contribution of an independent consultant if needed. All patient data that were available on the day of the decision were evaluated. Details on the decision-making process and on implementation of the decision were recorded. Additional information is provided in the
FOLLOW-UP AND OUTCOMES
All surviving patients were followed until day 90 after randomization. The primary outcome was survival with a favorable day-90 neurologic outcome, which was assessed with the use of the Cerebral Performance Category (CPC) scale (scores range from 1 to 5, with higher scores indicating greater disability). For this trial, a favorable neurologic outcome was defined as a CPC score of 1 (good cerebral performance or minor disability) or 2 (moderate disability).
The CPC score at 90 days was assessed during a semistructured telephone interview adapted from the validated French version
of the 5-item Glasgow Outcome Scale
by a single psychologist who was unaware of the group assignments and who had been specifically trained for the trial.
The secondary outcomes were mortality, mechanical ventilation duration, length of stay in the ICU and hospital, infections, and hematologic adverse events.
We assumed that 23% of the patients in the hypothermia group and 14% of those in the normothermia group would have a CPC score of 1 or 2 on day 90.
We calculated that the trial would need to enroll 584 patients in order for the trial to have 80% power to detect an absolute between-group difference of 9 percentage points in the percentage of patients with the primary outcome, at a two-sided significance level of 5%, and given that two interim analyses were planned.
The results of two interim analyses, performed after the inclusion of 200 and 400 patients, were provided to an independent data and safety monitoring board but not to the investigators. The Peto and Haybittle rule was applied, with the significance level set at 0.001 for both interim analyses, and the significance level associated with the final analysis was set at 0.049 to maintain an overall type I error of 5%.
Categorical variables were described with the use of numbers and percentages, and continuous variables with the use of medians and interquartile ranges. To estimate the treatment effect on the primary outcome, we estimated the between-group difference and its 95% confidence interval using a linear model with an identity link function. Missing data were handled by assuming that patients with missing data had died. To assess the consistency of the treatment effect on the primary outcome across prespecified subgroups, differences in percentages and 95% confidence intervals across subgroups were assessed by linear models with identity link functions including interaction terms. An additional analysis was performed with respect to the primary outcome and was adjusted for stratification variables (trial center and cardiac vs. noncardiac cause of arrest) with the use of a linear model with an identity link function.
Secondary outcomes that were expressed as proportions with their 95% confidence intervals were compared between groups. Secondary outcomes that were reported as cumulative incidences were analyzed with the use of the competing-risks approach, with death, ICU discharge, and hospital discharge as the competing risks. Confidence intervals for the secondary efficacy end points were not adjusted for multiplicity, and therefore inferences drawn from these intervals may not be reproducible. We used SAS software, version 9.4 (SAS Institute), and R software, version 3.3.1, for the statistical analyses.