The n e w e ng l a n d j o u r na l of m e dic i n e

Original Article

Hypothermia versus Normothermia

after Out-of-Hospital Cardiac Arrest
J. Dankiewicz, T. Cronberg, G. Lilja, J.C. Jakobsen, H. Levin, S. Ullén, C. Rylander,
M.P. Wise, M. Oddo, A. Cariou, J. Bělohlávek, J. Hovdenes, M. Saxena,
H. Kirkegaard, P.J. Young, P. Pelosi, C. Storm, F.S. Taccone, M. Joannidis,
C. Callaway, G.M. Eastwood, M.P.G. Morgan, P. Nordberg, D. Erlinge, A.D. Nichol,
M.S. Chew, J. Hollenberg, M. Thomas, J. Bewley, K. Sweet, A.M. Grejs,
S. Christensen, M. Haenggi, A. Levis, A. Lundin, J. Düring, S. Schmidbauer,
T.R. Keeble, G.V. Karamasis, C. Schrag, E. Faessler, O. Smid, M. Otáhal,
M. Maggiorini, P.D. Wendel Garcia, P. Jaubert, J.M. Cole, M. Solar, O. Borgquist,
C. Leithner, S. Abed‑Maillard, L. Navarra, M. Annborn, J. Undén, I. Brunetti,
A. Awad, P. McGuigan, R. Bjørkholt Olsen, T. Cassina, P. Vignon, H. Langeland,
T. Lange, H. Friberg, and N. Nielsen, for the TTM2 Trial Investigators*​​

A BS T R AC T

BACKGROUND
Targeted temperature management is recommended for patients after cardiac arrest, The authors’ full names, academic de‑
but the supporting evidence is of low certainty. grees, and affiliations are listed in the
Appendix. Address reprint requests to
METHODS Dr. Nielsen at the Department of Anes‑
In an open-label trial with blinded assessment of outcomes, we randomly assigned thesiology and Intensive Care, Intensive
Care Unit, Helsingborg Hospital, S Vall‑
1900 adults with coma who had had an out-of-hospital cardiac arrest of presumed gatan 5, 251 87, Helsingborg, Sweden, or
cardiac or unknown cause to undergo targeted hypothermia at 33°C, followed by at ­niklas​.­nielsen@​­med​.­lu​.­se.
controlled rewarming, or targeted normothermia with early treatment of fever *A complete list of TTM2 Trial Investiga‑
(body temperature, ≥37.8°C). The primary outcome was death from any cause at 6 tors is provided in the Supplementary
months. Secondary outcomes included functional outcome at 6 months as as- Appendix, available at NEJM.org.

sessed with the modified Rankin scale. Prespecified subgroups were defined ac- N Engl J Med 2021;384:2283-94.
cording to sex, age, initial cardiac rhythm, time to return of spontaneous circula- DOI: 10.1056/NEJMoa2100591
Copyright © 2021 Massachusetts Medical Society.
tion, and presence or absence of shock on admission. Prespecified adverse events
were pneumonia, sepsis, bleeding, arrhythmia resulting in hemodynamic compro-
mise, and skin complications related to the temperature management device.
RESULTS
A total of 1850 patients were evaluated for the primary outcome. At 6 months, 465
of 925 patients (50%) in the hypothermia group had died, as compared with 446 of
925 (48%) in the normothermia group (relative risk with hypothermia, 1.04; 95%
confidence interval [CI], 0.94 to 1.14; P = 0.37). Of the 1747 patients in whom the
functional outcome was assessed, 488 of 881 (55%) in the hypothermia group had
moderately severe disability or worse (modified Rankin scale score ≥4), as compared
with 479 of 866 (55%) in the normothermia group (relative risk with hypothermia,
1.00; 95% CI, 0.92 to 1.09). Outcomes were consistent in the prespecified subgroups.
Arrhythmia resulting in hemodynamic compromise was more common in the hypo-
thermia group than in the normothermia group (24% vs. 17%, P<0.001). The inci-
dence of other adverse events did not differ significantly between the two groups.
CONCLUSIONS
In patients with coma after out-of-hospital cardiac arrest, targeted hypothermia did
not lead to a lower incidence of death by 6 months than targeted normothermia.
(Funded by the Swedish Research Council and others; TTM2 ClinicalTrials.gov num-
ber, NCT02908308.)
n engl j med 384;24  nejm.org  June 17, 2021 2283
The New England Journal of Medicine
Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

I
nternational guidelines recommend mittees in each participating country. Written
targeted temperature management to prevent informed consent was waived, deferred, or ob-
hypoxic–ischemic brain damage in patients tained from a legal surrogate, depending on the
with coma after cardiac arrest.1,2 The evidence to circumstances, and was obtained from each pa-
support these recommendations originated in tient who regained mental capacity. An indepen-
trials involving patients who had been resusci- dent data and safety monitoring committee re-
tated from out-of-hospital cardiac arrest of a viewed the data and performed one prespecified,
presumed cardiac cause and shockable initial blinded interim analysis. Additional details on
rhythms.3,4 These trials suggested an increased the trial design, including investigator responsi-
survival and improved neurologic outcome in bilities, are described in the Supplementary Ap-
patients who underwent hypothermia at 33°C. pendix, available at NEJM.org. There was no
A recent trial involving patients who had cardiac commercial funding for the trial.
arrest with nonshockable rhythm showed better
neurologic outcomes with targeted hypothermia Patients
at 33°C than with targeted normothermia at We consecutively screened adults (≥18 years of
37°C.5 Trials comparing the level of targeted age) who had been admitted to the hospital after
temperature management (33°C or 36°C) and the out-of-hospital cardiac arrest of a presumed car-
duration of this management (24 hours or 48 diac or unknown cause, irrespective of the initial
hours) have not indicated a dose effect.6,7 rhythm. All the patients were unconscious and
Although guidelines strongly recommend tar- not able to obey verbal commands (score of <4 on
geted temperature management with a constant the Full Outline of Unresponsiveness [FOUR]
target between 32°C and 36°C, they also state scale,12 which ranges from 0 to 4, with higher
that the overall evidence is of low certainty. A scores indicating better motor function) and did
systematic review that included a meta-analysis not have a verbal response to pain. Eligible patients
and trial sequential analysis indicated that the had more than 20 consecutive minutes of spon-
available trials had high risks of bias and ran- taneous circulation after resuscitation.13 The main
dom errors.8 exclusion criteria were an interval from return of
Fever has been proposed as a risk factor for spontaneous circulation to screening of more
an unfavorable neurologic outcome after cardiac than 180 minutes, unwitnessed cardiac arrest with
arrest, although it is unknown whether there is asystole as the initial rhythm, and limitations in
a causal and modifiable relationship.9 Accord- care. Detailed inclusion and exclusion criteria
ingly, we conducted the randomized Targeted are provided in the Supplementary Appendix.
Hypothermia versus Targeted Normothermia af-
ter Out-of-Hospital Cardiac Arrest (TTM2) trial Randomization and Blinding
to assess the beneficial and harmful effects of After eligibility screening, patients were random­
hypothermia as compared with normothermia ly assigned in a 1:1 ratio to undergo hypother-
and early treatment of fever in patients after mia or normothermia. Randomization was per-
cardiac arrest. We hypothesized that at 6 months, formed with the use of a Web-based system
the incidence of death would be lower in the involving permuted blocks of varying sizes and
hypothermia group than in the normothermia was stratified according to trial site and coen-
group. rollment in the Targeted Therapeutic Mild Hyper-
capnia after Resuscitated Cardiac Arrest (TAME)
trial (ClinicalTrials.gov number, NCT03114033).
Me thods
Health professionals caring for the trial pa-
Trial Design tients were aware of the trial-group assignments
The design of this international, investigator- because of inherent problems with blinding body
initiated superiority trial and its statistical analy- temperature. The physicians assessing neuro-
sis plan have been published previously.10,11 The logic prognosis, assessors of functional outcome,
protocol (available with the full text of this article and study administrators were unaware of the
at NEJM.org) was approved by the ethics com- trial-group assignments. During the analysis

2284 n engl j med 384;24  nejm.org  June 17, 2021

The New England Journal of Medicine

Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Hypothermia or Normothermia after Cardiac Arrest

phase, the statisticians and authors were un- Primary and Secondary Outcomes
aware of the trial-group assignments, which The primary outcome was death from any cause
were identified as Y and Z. A manuscript was at 6 months. The main secondary outcome was
written for each scenario before the randomiza- a poor functional outcome at 6 months, defined
tion code was broken.14 as a score of 4 to 6 on the modified Rankin
scale.15,16 Scores on the modified Rankin scale
Trial Intervention range from 0 to 6, with 0 representing no symp-
The intervention period of 40 hours began at the toms, 1 no clinically significant disability, 2 slight
time of randomization. Patients who were as- disability, 3 moderate disability, 4 moderately
signed to undergo hypothermia were immedi- severe disability, 5 severe disability, and 6 death.
ately cooled with a surface or intravascular A trained outcome assessor used a structured
temperature-management device to a target tem- questionnaire to evaluate the patient’s condition.
perature of 33°C. This target was maintained The functional score was determined after face-
until 28 hours after randomization, followed by to-face or telephone interviews with patients,
rewarming to 37°C in hourly increments of one relatives, and health care providers.17
third of a degree. In the normothermia group, If a structured assessment could not be com-
the aim was to maintain a temperature of 37.5°C pleted, a binary assessment based on all avail-
or less. If conservative and pharmacologic mea- able data (including medical records) was per-
sures were insufficient and the body temperature formed; functional outcome was classified as
reached 37.8°C or higher, cooling with a surface “good” or “poor” on the basis of a dichotomized
or intravascular temperature-management device modified Rankin scale (a score of 0 to 3 or 4 to 6).
was initiated with a target temperature of 37.5°C. This post hoc approach was used because of the
No active warming or cooling was provided for restrictions imposed during the coronavirus dis-
patients in the normothermia group who had a ease 2019 pandemic.
spontaneous body temperature below 37.8°C. Other secondary outcomes were the number
Sedation was mandated in both groups until the of days the patient was alive and out of the hos-
end of the intervention period. After the inter- pital until day 180, survival determined in a
vention period, a normothermic target (36.5°C time-to-death analysis, and health-related qual-
to 37.7°C) was maintained until 72 hours after ity of life, which was assessed with the use of
randomization in patients who remained sedat- the visual-analogue scale on the European Qual-
ed or comatose. Details of the trial interventions ity of Life–5-Dimension–5-Level questionnaire
are provided in the Supplementary Appendix. (EQ-5D-5L), which ranges from 0 to 100, with
higher scores indicating better health status as
Assessment of Neurologic Prognosis assessed by the patient.15 Verification of trial
and Withdrawal of Life Support data and the outcome measures are described in
At 96 hours after randomization or later, a phy- the Supplementary Appendix.
sician who was unaware of the intervention as-
signments performed a neurologic assessment Adverse Events
of patients who remained in the intensive care Prespecified adverse events were pneumonia, sep-
unit (ICU). The physician assessed whether the sis, bleeding, arrhythmia resulting in hemody-
criteria for a likely poor neurologic outcome namic compromise, and skin complications re-
were present. lated to the device used for targeted temperature
All decisions about withdrawal of life-sus- management. Definitions of these adverse events
taining therapy were at the discretion of the are provided in the Supplementary Appendix.
treating physician, guided by the protocol. After
assessment of neurologic prognosis, withdrawal Statistical Analysis
of life-sustaining therapies due to a presumed We estimated that a sample of 1862 patients
poor neurologic prognosis was allowed. (Details would provide 90% power to detect a relative
regarding the neurologic evaluation are provided reduction of 15% in the risk of death in the hypo-
in the Supplementary Appendix.) thermia group, as compared with the normo-

n engl j med 384;24  nejm.org  June 17, 2021 2285

The New England Journal of Medicine
Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 1. Baseline Characteristics of the Intention-to-Treat Population.*

Hypothermia Normothermia
Characteristic (N = 930) (N = 931)
Demographic characteristics
Age — yr 64±13 63±14
Male sex — no. (%) 742 (80) 735 (79)
Medical history
Hypertension — no. (%) 345 (37) 298 (32)
Diabetes — no. (%) 173 (19) 167 (18)
Myocardial infarction — no. (%) 139 (15) 154 (17)
PCI — no. (%) 130 (14) 140 (15)
Coronary-artery bypass grafting — no. (%) 73 (8) 76 (8)
Heart failure — no. (%) 90 (10) 93 (10)
NYHA III or IV heart failure — no./total no. (%)† 20/906 (2) 23/904 (3)
Median Charlson comorbidity index (IQR)‡ 3 (2–4) 3 (1–4)
Characteristics of the cardiac arrest — no. (%)
Location at cardiac arrest
Place of residence 487 (52) 491 (53)
Public place 338 (36) 320 (34)
Other 105 (11) 120 (13)
Bystander-witnessed cardiac arrest 850 (91) 852 (92)
Bystander-performed CPR 759 (82) 728 (78)
First monitored rhythm — no. (%)
Shockable rhythm 671 (72) 700 (75)
Ventricular fibrillation 576 (62) 585 (63)
Nonperfusing ventricular tachycardia 31 (3) 29 (3)
ROSC after bystander-initiated defibrillation 24 (3) 41 (4)
Unknown rhythm, shock administered 40 (4) 45 (5)
Nonshockable rhythm 259 (28) 231 (25)
Pulseless electrical activity 117 (13) 113 (12)
Asystole 124 (13) 100 (11)
Unknown rhythm, no shock administered 18 (2) 18 (2)
Median time from cardiac arrest to sustained ROSC (IQR) — min§ 25 (16–40) 25 (17–40)
Median time from cardiac arrest to randomization — min (IQR) 136 (103–170) 133 (99–173)
Clinical characteristics on admission
Tympanic temperature — °C¶ 35.3±1.1 35.4±1.1
FOUR motor score‖ 0 0
Bilateral corneal reflexes present — no./total no. (%) 168/511 (33) 194/537 (36)
Bilateral pupillary reflexes present — no./total no. (%) 535/761 (70) 529/776 (68)
Arterial pH** 7.2±0.2 7.2±0.2
Arterial lactate level — mmol/liter†† 5.9±4.4 5.8±4.2
Shock — no. (%)‡‡ 261 (28) 275 (30)
ST-segment elevation myocardial infarction — no./total no. (%) 379/918 (41) 370/921 (40)

2286 n engl j med 384;24  nejm.org  June 17, 2021

The New England Journal of Medicine

Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Hypothermia or Normothermia after Cardiac Arrest

Table 1. (Continued.)

* Plus–minus values are means ±SD. CPR denotes cardiopulmonary resuscitation, IQR interquartile range, PCI percuta‑
neous coronary intervention, and ROSC return of spontaneous circulation.
† New York Heart Association (NYHA) heart failure class was not assessed in 51 patients (24 in the hypothermia group
and 27 in the normothermia group) who had a history of heart failure.
‡ On the Charlson comorbidity index, each comorbidity category is weighted from 1 to 6 on the basis of adjusted risk
of death or resource use, and the sum of the weights produces the score. A score of 0 indicates an absence of known
coexisting conditions, and higher scores indicate higher risks of death and greater resource use.
§ For unwitnessed cardiac arrests, the time to ROSC was calculated from the time of the emergency call.
¶ Tympanic temperature was assessed in 1559 patients.
‖ Full Outline of Unresponsiveness (FOUR) motor scores range from 0 to 4, with higher scores indicating better motor
function. The FOUR motor score was assessed in 1696 patients.
** Arterial pH was measured in 1829 patients.
†† The arterial lactate level was measured in 1781 patients.
‡‡ Shock at admission was defined as a systolic blood pressure of less than 90 mm Hg for more than 30 minutes or
end-organ hypoperfusion (cool arms and legs, urine output <30 ml per hour, and heart rate <60 beats per minute).

thermia group, at a two-sided alpha level of 0.05 and 2 patients underwent randomization twice,
(absolute risk reduction of 7.5 percentage points). resulting in an intention-to-treat population of
The estimated relative risk was based on results 1861, of whom 930 were assigned to the hypo-
from earlier trials of hypothermia for cardiac ar- thermia group and 931 to the normothermia
rest.10 To allow for loss to follow-up and with- group (Fig. S1 in the Supplementary Appendix).
drawn consent, a sample size of 1900 was chosen. Baseline characteristics are reported in Table 1.
The principal trial analyses were performed in Details regarding procedures and administered
the intention-to-treat population, defined as all drugs, assessment of neurologic prognosis, with-
randomly assigned patients except those for drawal of life-sustaining therapy, length of ICU
whom consent was withdrawn. Dichotomous out- and hospital stay, and data regarding coenroll-
comes, including the primary analysis, were ment in the TAME trial are provided in Tables S1
assessed with the use of a mixed-effects general- through S6 and Figures S2 and S3.
ized linear model with a logit link with adjust-
ment for stratification variables and were report- Temperature Intervention
ed as population-level (marginal) relative risks The temperature curves are shown in Figure 1.
derived by G-computation. Analysis of survival In the hypothermia group, the median time from
data was performed with Cox regression. For all the start of the intervention until a temperature
regression analyses, we tested for an interactionof 34°C was reached was 3 hours. In this group,
effect between group assignment and assign- 53 of 930 patients (6%) were rewarmed before
ment in the TAME trial. We made no assump- 40 hours after randomization, as allowed by the
tions regarding the pattern of missing data, protocol, primarily because of cardiovascular
which were handled according to the statistical instability and arrhythmias (Table S7). A total of
analysis plan.11,18 A P value of less than 0.05 was
882 of 930 patients (95%) in the hypothermia
considered to indicate statistical significance for
group and 428 of 931 patients (46%) in the nor-
the primary outcome. Secondary outcomes are mothermia group received cooling with a device.
presented with 95% confidence intervals and Among patients who received cooling, the types
were not adjusted for multiplicity. All analyses of devices used in each treatment group were
were performed with the use of R: A Language similar (70% surface and 30% intravascular in
and Environment for Statistical Computing.19 the hypothermia group and 69% surface and
31% intravascular in the normothermia group).
In the hypothermia group, the reasons for not
R e sult s
receiving cooling with a device were intracranial
Patients hemorrhage, early death, early awakening, hemo-
A total of 1900 patients were enrolled between dynamic instability, and referral for cardiac
November 2017 and January 2020. Consent could surgery, whereas the main reason in the normo-
not be obtained or was withdrawn in 37 patients, thermia group was not reaching the threshold

n engl j med 384;24  nejm.org  June 17, 2021 2287

The New England Journal of Medicine
Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Functional outcome was assessed according

Normothermia group Hypothermia group
to the modified Rankin scale in 1747 of 1861
39
patients (94%) (Fig. S8). A structured assessment
was performed in a face-to-face interview (72%),
38
by phone (23%), or by proxy interview (5%). In
addition, functional outcome was classified only
37
as “good” or “poor” on the basis of telephone
interviews with relatives and health care provid-
Body Temperature (°C)

36
ers and on the basis of medical records in 37 of
930 patients (4%) in the hypothermia group and
35
45 of 931 patients (5%) in the normothermia
group. In total, functional outcome was assessed
34
in 1829 of 1861 patients (98%).
At 6 months, 488 of 881 patients (55%) in the
33
hypothermia group and 479 of 866 patients
(55%) in the normothermia group had a modi-
32
fied Rankin scale score of 4 to 6 (relative risk
with hypothermia, 1.00; 95% CI, 0.92 to 1.09).
31
0 10 20 30 40 In the binary assessment of functional outcome,
Hours since Randomization
495 of 918 patients (54%) in the hypothermia
group and 493 of 911 patients (54%) in the nor-
Figure 1. Body Temperature during the Intervention Period. mothermia group had a poor functional out-
Shown are body-temperature curves in the hypothermia and normothermia come (relative risk in the hypothermia group,
groups for the patients in whom a bladder temperature was recorded. The 1.00; 95% CI, 0.91 to 1.08). The effect of the
median number of temperature recordings was 38 in both the hypothermia temperature intervention on functional outcome
group and the normothermia group, out of 41 possible recordings. The tem‑
perature curves show the means, and the I bars indicate ±2 SD (95% of the
was consistent across the prespecified subgroups
observations are within the error bars). The median time from cardiac arrest (Fig. 2B).
to randomization in the trial was 135 minutes. Health-related quality of life as assessed with
the use of the EQ-5D-5L visual-analogue scale
was similar in the hypothermia and normother-
for fever. Additional data regarding tempera- mia groups, regardless of whether the patients
tures and shivering are available in Figures S4 who died were included (with the score on the
through S7 and Table S8. EQ-5D-5L visual-analogue scale set to 0) or only
those who survived were assessed (mean between-
Primary and Secondary Outcomes group difference in patients who survived to
Data on the primary outcome were missing for 6 months, −0.8 points; 95% CI, −3.6 to 2.0)
11 patients (5 in the hypothermia group and 6 in (Table S9). The distribution of days when the
the normothermia group) of 1861 patients over- patients were alive and out of the hospital was
all (<1%). At 6 months, 465 of 925 patients (50%) similar in the two groups (Fig. S9).
in the hypothermia group and 446 of 925 pa- Best–worst and worst–best analyses indicated
tients (48%) in the normothermia group had that missing data did not have the potential to
died (relative risk with hypothermia, 1.04; 95% affect the results of the analyses of both death
confidence interval [CI], 0.94 to 1.14; P = 0.37). from any cause and functional outcome (Table
The effect of the temperature intervention on S10). Additional sensitivity analyses are reported
death at 6 months was consistent across the in Table S11. There were no significant interac-
prespecified subgroups (Fig. 2A) and when as- tions between group assignments in the current
sessed in a time-to-event analysis (hazard ratio trial and assignments in the TAME trial for any
in the hypothermia group, 1.08; 95% CI, 0.95 to of the outcomes (range of P for interaction, 0.58
1.23) (Fig. 3). to 0.94) (Table S12).

2288 n engl j med 384;24  nejm.org  June 17, 2021

The New England Journal of Medicine

Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Hypothermia or Normothermia after Cardiac Arrest

A Death at 6 Months
Subgroup Hypothermia Normothermia Relative Risk of Death (95% CI)
no. of patients
All patients 925 925 1.04 (0.94–1.14)
Sex
Male 738 729 1.03 (0.92–1.15)
Female 187 196 1.10 (0.94–1.29)
Age
<65 yr 421 457 0.99 (0.83–1.18)
≥65 yr 504 468 1.04 (0.94–1.15)
Time to ROSC from cardiac arrest
<25 min 419 416 1.09 (0.91–1.33)
≥25 min 506 509 1.02 (0.92–1.12)
Initial rhythm
Nonshockable 259 231 1.04 (0.94–1.14)
Shockable 666 694 1.00 (0.87–1.15)
Shock on admission
Not present 665 651 1.07 (0.95–1.23)
Present 260 274 1.01 (0.89–1.15)
0.50 0.75 1.00 1.25 1.50

Hypothermia Better Normothermia Better

B Modified Rankin Scale Score of 4–6 at 6 Months

Subgroup Hypothermia Normothermia Relative Risk of Score of 4–6 (95% CI)
no. of patients
All patients 881 866 1.00 (0.92–1.09)
Sex
Male 701 679 1.00 (0.90–1.10)
Female 180 187 1.03 (0.90–1.19)
Age
<65 yr 391 429 0.94 (0.79–1.10)
≥65 yr 490 437 1.01 (0.92–1.10)
Time to ROSC from cardiac arrest
<25 min 395 389 1.04 (0.87–1.24)
≥25 min 486 477 0.98 (0.90–1.07)
Initial rhythm
Nonshockable 252 218 1.00 (0.93–1.08)
Shockable 629 648 0.96 (0.84–1.08)
Shock on admission
Not present 629 606 1.03 (0.92–1.16)
Present 252 260 0.97 (0.86–1.08)
0.50 0.75 1.00 1.25 1.50

Hypothermia Better Normothermia Better

Figure 2. Subgroup Analysis of Death from Any Cause and the Modified Rankin Scale Score at 6 Months.
Shown are the results of the analyses of the primary outcome (death from any cause at 6 months) (Panel A) and of
the secondary outcome of a score of 4 to 6 on the modified Rankin scale (Panel B) in prespecified subgroups. Mod­
ified Rankin scale scores range from 0 to 6, with 0 representing no symptoms, 1 no clinically significant disability,
2 slight disability, 3 moderate disability, 4 moderately severe disability, 5 severe disability, and 6 death. Relative risks
are derived from a stratified generalized linear model with trial site as a random intercept. The forest plot shows the
relative risks for five prespecified subgroups. The horizontal bars represent 95% confidence intervals. The events
are the total events 6 months after randomization. For unwitnessed cardiac arrests, the time until a return of spon‑
taneous circulation (ROSC) was calculated from the time of the emergency call. Shock on admission was defined as
a systolic blood pressure of less than 90 mm Hg for more than 30 minutes or end-organ hypoperfusion (cool arms
and legs, urine output <30 ml per hour, and heart rate <60 beats per minute).

n engl j med 384;24  nejm.org  June 17, 2021 2289

The New England Journal of Medicine
Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

between the groups was similar, as was health-

1.00 related quality of life. The results were consis-
tent in the analysis of survival and in prespeci-
fied subgroups.
Probability of Survival

0.75
Our results contrast with findings of prac-
Normothermia tice-changing trials published in 2002 in which
0.50
Hypothermia a benefit of hypothermia was reported.3,4 Since
then, there have been changes in standards of
0.25 intensive care that may have influenced interven-
tion effects.20,21 Other explanations would be a
0.00 lower risk of bias in the current trial22 and a lower
0 40 80 120 160 risk of random error with a sample size that was
Days since Randomization five times the combined enrollment of the ear-
No. at Risk lier trials.23,24 Although the patient population
Normothermia
Hypothermia
925
925
506
474
491
468
484
462
480
461
we studied differed somewhat from those in
previous trials, our subgroup analyses indicate
Figure 3. Probability of Survival until 180 Days after Randomization.
that different eligibility criteria are unlikely to
Shown are Kaplan–Meier estimates of the probability of survival until 180
explain the discordance.
days after randomization among patients assigned to undergo hypothermia Our findings are consistent with those of a
or normothermia. Data are for the 1850 patients for whom survival status recent trial in which hypothermia at 33°C, as
(including time of death) was available. Data were censored according to compared with normothermia at 37°C, in patients
the last day of follow-up. with nonshockable rhythms was not shown to
reduce mortality.5 That trial indicated that hypo-
thermia may improve functional outcomes, but
Adverse Events this finding was based on a small number of
Prespecified adverse events are reported in Ta- events and was not replicated in the subgroup
ble 2. Arrhythmias resulting in hemodynamic of patients with initial nonshockable rhythm in
compromise were more common in the hypo- our trial.
thermia group than in the normothermia group The results of the current trial are broadly
(in 24% vs. 17%; P<0.001). There were no signifi- consistent with the results of our previous TTM
cant differences in other prespecified adverse (Target Temperature Management 33°C versus
events. Two unexpected serious, possibly inter- 36°C after Out-of-Hospital Cardiac Arrest) trial.6
vention-related adverse events occurred in each The combined results of the two trials imply a
group: an intravascular device–related thrombo- low likelihood of any meaningful clinical im-
sis in one patient in the hypothermia group and provement with hypothermia as compared with
two patients in the normothermia group, and normothermia, since 36°C may be considered to
bradycardia with worsening hemodynamic func- be the lower boundary of normothermia.
tion in one patient in the hypothermia group It is physiologically plausible that the interval
(see the Supplementary Appendix). between a cardiac event and the initiation of
hypothermia is related to potential benefits of
the intervention, a hypothesis that is supported
Discussion
by experiments in animals.25 In our trial, patients
In this randomized trial, we compared hypother- were cooled at a similar or faster rate than that
mia with normothermia in patients with coma in most previous trials.3,5-7 Since all participating
who had been resuscitated after out-of-hospital sites in our trial had previous experience with
cardiac arrest of a presumed cardiac or unknown the use of hypothermia, and a large percentage
cause. There was no significant difference be- of the patients in our trial underwent randomiza-
tween the two groups with respect to death and tion at cardiac arrest centers, the cooling rates
poor functional outcome at 6 months. The dis- we observed were probably faster than those that
tribution of scores on the modified Rankin scale are feasible in current clinical practice.

2290 n engl j med 384;24  nejm.org  June 17, 2021

The New England Journal of Medicine

Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Hypothermia or Normothermia after Cardiac Arrest

Table 2. Outcomes and Adverse Events.

Hypothermia Normothermia Relative Risk

Outcome or Event (N = 930) (N = 931) (95% CI)* P Value
Primary outcome: death from any cause at 6 mo 465/925 (50) 446/925 (48) 1.04 (0.94–1.14) 0.37
— no./total no. (%)
Main secondary outcome — no./total no. (%)
Score of 4–6 on modified Rankin scale at 6-mo 488/881 (55) 479/866 (55) 1.00 (0.92–1.09)
follow-up†
Poor functional outcome at 6 mo‡ 495/918 (54) 493/911 (54) 1.00 (0.91–1.08)
Score on modified Rankin scale at 6-mo follow-up
— no./total no. (%)†
0 140/881 (16) 148/866 (17)
1 87/881 (10) 80/866 (9)
2 132/881 (15) 127/866 (15)
3 34/881 (4) 32/866 (4)
4 16/881 (2) 20/866 (2)
5 7/881 (1) 13/866 (2)
6 465/881 (53) 446/866 (52)
Serious adverse events — no./total no. (%)
Arrhythmia resulting in hemodynamic com‑ 222/927 (24) 152/921 (16) 1.45 (1.21–1.75) <0.001
promise
Bleeding 44/927 (5) 46/922 (5) 0.95 (0.63–1.42) 0.81
Skin complication related to device used for 10/927 (1) 5/922 (<1) 1.99 (0.71–6.37) 0.21
targeted temperature management
Pneumonia 330/927 (36) 322/921 (35) 1.02 (0.90–1.15) 0.75
Sepsis 99/926 (11) 83/922 (9) 1.19 (0.90–1.57) 0.23

* The relative risks of death from any cause, a modified Rankin scale score of 4 to 6, and poor neurologic function at 6 months were adjusted
for the stratification variables. The relative risks of serious adverse events were adjusted for coenrollment status in the Targeted Therapeutic
Mild Hypercapnia after Resuscitated Cardiac Arrest (TAME) trial, but not for site. The relative risks of skin complications related to the device
used for targeted temperature management were unadjusted. The widths of the confidence intervals have not been adjusted for multiple
testing, so the intervals should not be used to infer definitive differences between the groups.
† Scores on the modified Rankin scale range from 0 to 6, with 0 representing no symptoms, 1 no clinically significant disability, 2 slight dis‑
ability, 3 moderate disability, 4 moderately severe disability, 5 severe disability, and 6 death. These results are based on data from a struc‑
tured interview
‡ These results are based on all available data..

Hypothermia did not increase the frequency chanical ventilation, were therefore included in
of pneumonia, sepsis, or bleeding, but arrhyth- the trial protocol in a form that was not neces-
mias causing hemodynamic compromise were sarily representative of clinical practice. It is
more common in the hypothermia group than unclear what influence these elements had on
in the normothermia group. Possible reasons for the outcomes. The trial also included a conserva-
this include electrolyte disturbances, fluid sta- tive protocol for assessment of neurologic prog-
tus, and a temperature effect on cardiac myo- nosis and guidance for withdrawal of life sup-
cytes.26 port, which may have influenced outcomes.
Our trial has several limitations. First, to iso- Second, staff members in the ICU were aware of
late the effect of hypothermia, both trial groups the assigned target temperature during the ICU
were treated similarly, except for the tempera- stay. We aimed to minimize this problem by us-
ture intervention. Elements of standard care in ing outcomes with a low risk of bias, outcome
the ICU, such as sedation, paralysis, and me- assessors who were unaware of the trial-group

n engl j med 384;24  nejm.org  June 17, 2021 2291

The New England Journal of Medicine
Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

assignments, and a conservative protocol for on the modified Rankin scale), and patient-­
determination of the neurologic prognosis and reported health-related quality of life (as mea-
withdrawal of life-supporting therapies. During sured on the EQ-5D-5L visual-analogue scale).
the analysis and writing process, the investiga- The large sample size, broad eligibility criteria,
tors, statisticians, and authors were unaware of and numerous hospitals and countries repre-
the temperature-group assignments, and writing sented in this trial increase the generalizability.
of the manuscript was performed in duplicate, Patients with coma after out-of-hospital cardi-
with the groups interchanged. Third, since we ac arrest who were treated with hypothermia did
did not include a control group without tempera- not have a lower incidence of death at 6 months
ture management, this trial leaves a knowledge than those who were treated with normothermia.
gap regarding whether any temperature man- Supported by independent research grants from nonprofit or
agement is better than no temperature manage- governmental agencies (the Swedish Research Council [Veten-
skapsrådet], Swedish Heart–Lung Foundation, Stig and Ragna
ment. Nonetheless, actual temperatures in the Gorthon Foundation, Knutsson Foundation, Laerdal Founda-
normothermia group were broadly similar to tion, Hans-Gabriel and Alice Trolle-Wachtmeister Foundation
those recorded in the control group of the Hypo- for Medical Research, and Regional Research Support in Region
Skåne) and by governmental funding of clinical research within
thermia after Cardiac Arrest trial, in which no the Swedish National Health Service.
temperature management was used.3 As com- Dr. Oddo reports receiving advisory board fees from Neu-
pared with that trial, about half the patients in roptics; Dr. Cariou, receiving lecture fees from Bard Medical;
Dr. Saxena, receiving consulting fees from Bard Medical; Dr.
the normothermia group in our trial were cooled Young, receiving lecture fees from Bard Medical; Dr. Storm,
with a device. Whether this type of fever control receiving lecture fees and travel support from Bard Medical and
is of benefit must be addressed in a separate consulting fees from BrainCool, Pfizer, Sedana Medical, and
Xenios; Dr. Taccone, receiving grant support from Bard Medical
trial. Fourth, concomitant care, except for seda- and ZOLL Medical; Dr. Callaway, holding patent US7269454B2
tion and prognostication, was not part of the on methods and devices to guide therapy for ventricular fibril-
protocol and was left to the discretion of partici- lation with waveform analysis, patent WO2016168181A1 on
identification of non-ST elevation ischemic events on electrocar-
pating hospitals. However, sites were instructed diography, and patent US6438419B1 on a method and apparatus
to treat the groups similarly, and the stratifica- using a scaling exponent for selective defibrillation; Dr. Mor-
tion for participating hospitals should have bal- gan, receiving fees for coordinating an educational workshop
from Bard Medical; Dr. Nichol, receiving grant support, paid
anced intersite differences. Fifth, the trial was to University College Dublin, from AM Pharma and grant sup-
limited to out-of-hospital cardiac arrest of a port, paid to Monash University, from Baxter Healthcare; Dr.
Chew, receiving lecture fees from Edwards Lifesciences; Dr.
presumed cardiac or unknown cause, so the Levis, receiving donated supplies from Prytime Medical Devices;
results are not fully applicable to other presenta- Dr. Keeble, receiving travel support from Zoll Circulation; Dr.
tions of cardiac arrest. However, a lack of cere- Leithner, receiving lecture fees, paid to Charité Universitätz-
medizin, from Bard Medical and ZOLL Medical; Dr. Friberg,
bral perfusion is the primary cause of hypoxic– receiving fees for academic advising from TEQCool; and Dr.
ischemic encephalopathy in cardiac arrest, Nielsen, receiving lecture fees from Bard Medical and consult-
regardless of where the event occurs or the cause ing fees from BrainCool. No other potential conflict of interest
relevant to this article was reported.
of arrest. Finally, one fifth of the patients were Disclosure forms provided by the authors are available with
also enrolled in the TAME trial. We did not an- the full text of this article at NEJM.org.
ticipate any between-trial interaction, an expecta- A data sharing statement provided by the authors is available
with the full text of this article at NEJM.org.
tion that was supported by our analyses, although We thank the trial patients, their relatives, and clinical and
such comparisons were probably underpowered. research staff at all the trial sites; Audrey Shearer and col-
Our results were consistent across the objec- leagues at Spiral Software; Richard Frobell, Kristina Källén,
Ulf Malmqvist, Kathy Rowan, David Harrison, Paul Mouncey,
tive outcome of death from any cause, the clini- Manu Shankar-Hari, Duncan Young, Karolina Palmér, and Vik-
cian-reported functional outcome (as measured tor Hultqvist; and the TAME trial group.

Appendix
The authors’ full names and academic degrees are as follows: Josef Dankiewicz, M.D., Ph.D., Tobias Cronberg, M.D., Ph.D., Gisela
Lilja, O.T., Ph.D., Janus C. Jakobsen, M.D., Ph.D., Helena Levin, M.Sc., Susann Ullén, Ph.D., Christian Rylander, M.D., Ph.D., Matt P.
Wise, M.B., B.Ch., D.Phil., Mauro Oddo, M.D., Alain Cariou, M.D., Ph.D., Jan Bělohlávek, M.D., Ph.D., Jan Hovdenes, M.D., Ph.D.,
Manoj Saxena, M.B., B.Ch., Ph.D., Hans Kirkegaard, M.D., D.M.Sc., Paul J. Young, M.D., Ph.D., Paolo Pelosi, M.D., Christian Storm,
M.D., Ph.D., Fabio S. Taccone, M.D., Ph.D., Michael Joannidis, M.D., Clifton Callaway, M.D., Ph.D., Glenn M. Eastwood, R.N., Ph.D.,
Matt P.G. Morgan, M.B., B.Ch., Ph.D., Per Nordberg, M.D., Ph.D., David Erlinge, M.D., Ph.D., Alistair D. Nichol, Ph.D., Michelle S.
Chew, M.D., Ph.D., Jacob Hollenberg, M.D., Ph.D., Matthew Thomas, M.B., B.Ch., Jeremy Bewley, M.B., B.Ch., Katie Sweet, R.N.,
B.Sc., Anders M. Grejs, M.D., Ph.D., Steffen Christensen, M.D., Ph.D., Matthias Haenggi, M.D., Anja Levis, M.D., Andreas Lundin,
M.D., Joachim Düring, M.D., Simon Schmidbauer, M.D., Thomas R. Keeble, M.B., B.S., M.D., Grigoris V. Karamasis, M.D., Claudia

2292 n engl j med 384;24  nejm.org  June 17, 2021

The New England Journal of Medicine

Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Hypothermia or Normothermia after Cardiac Arrest

Schrag, M.D., Edith Faessler, M.D., Ondrej Smid, M.D., Ph.D., Michal Otáhal, M.D., Ph.D., Marco Maggiorini, M.D., Pedro D. Wendel
Garcia, M.Sc., Paul Jaubert, M.D., Jade M. Cole, R.N., B.Sc., Miroslav Solar, M.D., Ph.D., Ola Borgquist, M.D., Ph.D., Christoph
Leithner, M.D., Samia Abed‑Maillard, M.Sc., Leanlove Navarra, R.N., B.Sc., Martin Annborn, M.D., Ph.D., Johan Undén, M.D., Ph.D., Iole
Brunetti, M.D., Akil Awad, M.D., Peter McGuigan, M.B., B.Ch., Roy Bjørkholt Olsen, M.D., Ph.D., Tiziano Cassina, M.D., Ph.D.,
Philippe Vignon, M.D., Ph.D., Halvor Langeland, M.D., Theis Lange, Ph.D., Hans Friberg, M.D., Ph.D., and Niklas Nielsen, M.D., Ph.D.
The authors’ affiliations are as follows: the Department of Clinical Sciences Lund, Sections of Cardiology (J. Dankiewicz, D.E.),
Neurology (T. Cronberg, G.L.), and Anesthesiology and Intensive Care (H. Levin, O.B.), Skåne University Hospital Lund, Lund Univer-
sity and Clinical Studies Sweden — Forum South, Skåne University Hospital (S.U.), Lund; the Department of Clinical Sciences Lund,
Section of Anesthesia and Intensive Care, Skåne University Hospital Malmö, Malmö, (J. Düring, S.S., H.F.); the Department of Clinical
Sciences Lund, Sections of Anesthesiology and Intensive Care (M.A., N.N.) and Clinical Sciences Helsingborg (N.N.), Helsingborg
Hospital, Helsingborg; the Department of Clinical Sciences Lund, Section of Anesthesiology and Intensive Care Lund, Hallands Hospi-
tal, Halmstad (J.U.); the Department of Anesthesiology and Intensive Care Medicine, Institute of Clinical Sciences, Sahlgrenska Acad-
emy, University of Gothenburg, Gothenburg (C.R., A. Lundin); the Department of Clinical Science and Education, Center for Resuscita-
tion Science, Karolinska Institutet, Södersjukhuset, Stockholm (P.N., J. Hollenberg, A.A.); and the Department of Anesthesiology, In-
tensive Care, and Acute Medicine, Linköping University, Linköping (M.S.C.) — all in Sweden; Copenhagen Trial Unit, Center for
Clinical Intervention Research, Copenhagen University Hospital (J.C.J.), and the Section of Biostatistics, Faculty of Health and Medical
Sciences (T.L.), University of Copenhagen, Copenhagen, the Department of Regional Health Research, the Faculty of Health Sciences,
University of Southern Denmark, Odense (J.C.J.), the Research Center for Emergency Medicine, the Department of Clinical Medicine
(H.K.), and the Department of Intensive Care (A.M.G., S.C.), Aarhus University Hospital, Aarhus — all in Denmark; Adult Critical Care,
University Hospital of Wales, Cardiff (M.P.W., M.P.G.M., J.M.C.), the Department of Intensive Care, Bristol Royal Infirmary, Bristol
(M.T., J. Bewley, K.S.), Essex Cardiothoracic Centre, Basildon (T.R.K., G.V.K.), Anglia Ruskin University School of Medicine, Chelms-
ford, Essex (T.R.K., G.V.K.), and the Department of Anesthesiology and Intensive Care, Royal Victoria Hospital, Belfast (P.M.) — all in
the United Kingdom; Neuroscience Critical Care Research Group and Adult Intensive Care Medicine Service, Centre Hospitalier Univer-
sitaire Vaudois–Lausanne University Hospital and University of Lausanne, Lausanne (M. Oddo, S.A.-M.), the Departments of Intensive
Care Medicine (M.H.) and Anesthesiology and Pain Medicine, Inselspital (A. Levis), Bern University Hospital, University of Bern, Bern,
the Intensive Care Department, Kantonsspital St. Gallen, St. Gallen (C. Schrag, E.F.), the Institute of Intensive Care Medicine, Univer-
sity Hospital Zurich, Zurich (M.M., P.D.W.G.), and the Cardiac Anesthesia and Intensive Care Department, Instituto Cardiocentro Ticino,
Lugano (T. Cassina) — all in Switzerland; Descartes University of Paris and Cochin University Hospital, Paris (A.C., P.J.), Medical-Surgical
Intensive Care Unit, Dupuytren Teaching Hospital, Limoges (P.V.) — all in France; the 2nd Department of Medicine (J. Bělohlávek,
O.S.), and the Department of Anesthesiology and Intensive Care Medicine (M. Otáhal), General University Hospital and First Faculty of
Medicine, Charles University, Prague, the 1st Department of Internal Medicine–Cardioangiology, University Hospital Hradec Králové,
and Faculty of Medicine, Charles University, Hradec Králové (M. Solar) — all in the Czech Republic; the Department of Anesthesiology,
Division of Emergencies and Critical Care, Oslo University Hospital, Rikshospitalet, Oslo (J. Hovdenes), the Department of Anesthesiol-
ogy, Sørlandet Hospital, Arendal (R.B.O.), the Department of Anesthesiology and Intensive Care Medicine, St. Olav’s University Hospi-
tal, and the Department of Circulation and Medical Imaging, Faculty of Medicine and Health Sciences, Norwegian University of Science
and Technology, Trondheim (H. Langeland) — all in Norway; the Division of Critical Care and Trauma, George Institute for Global
Health, and Bankstown–Lidcombe Hospital, South Western Sydney Local Health District, Sydney (M. Saxena), and the Australian and
New Zealand Intensive Care Research Centre, School of Public Health and Preventive Medicine (G.M.E., A.D.N.), and the Department
of Intensive Care, Alfred Health (A.D.N.), Monash University, Melbourne — all in Australia; the Medical Research Institute of New
Zealand, Intensive Care Unit, Wellington Hospital, Wellington (P.J.Y., L.N.); the Departments of Surgical Sciences and Integrated Diag-
nostics (P.P.) and Anesthesiology and Intensive Care, San Martino Policlinico Hospital, IRCCS for Oncology and Neuroscience (P.P.,
I.B.), University of Genoa, Genoa, Italy; the Department of Nephrology and Medical Intensive Care (C. Storm), and Klinik und Hoch-
schulambulanz für Neurologie (C.L.), Charité Universitätzmedizin, Berlin, Germany; the Department of Intensive Care, Erasme Univer-
sity Hospital, Université Libre de Bruxelles, Brussels (F.S.T.); the Division of Intensive Care and Emergency Medicine, Department of
Internal Medicine, Medical University Innsbruck, Innsbruck, Austria (M.J.); the Department of Emergency Medicine, University of
Pittsburgh, Pittsburgh (C.C.); and University College Dublin Clinical Research Centre at St. Vincent’s University Hospital, Dublin, Ire-
land (A.D.N.).

References
1. Nolan JP, Soar J, Cariou A, et al. Euro- 4. Bernard SA, Gray TW, Buist MD, et al. 8. Nielsen N, Friberg H, Gluud C, Herlitz
pean Resuscitation Council and European Treatment of comatose survivors of out- J, Wetterslev J. Hypothermia after cardiac
Society of Intensive Care Medicine guide- of-hospital cardiac arrest with induced arrest should be further evaluated —
lines for post-resuscitation care 2015: hypothermia. N Engl J Med 2002;​346:​557- a systematic review of randomised trials
section 5 of the European Resuscitation 63. with meta-analysis and trial sequential
Council guidelines for resuscitation 2015. 5. Lascarrou J-B, Merdji H, Le Gouge A, analysis. Int J Cardiol 2011;​151:​333-41.
Resuscitation 2015;​95:​202-22. et al. Targeted temperature management 9. Zeiner A, Holzer M, Sterz F, et al. Hy-
2. Panchal AR, Bartos JA, Cabañas JG, for cardiac arrest with nonshockable perthermia after cardiac arrest is associ-
et al. Part 3: adult basic and advanced life rhythm. N Engl J Med 2019;​381:​2327-37. ated with an unfavorable neurologic out-
support: 2020 American Heart Associa- 6. Nielsen N, Wetterslev J, Cronberg T, come. Arch Intern Med 2001;​161:​2007-12.
tion guidelines for cardiopulmonary re- et al. Targeted temperature management 10. Dankiewicz J, Cronberg T, Lilja G, et al.
suscitation and emergency cardiovascular at 33°C versus 36°C after cardiac arrest. Targeted hypothermia versus targeted nor-
care. Circulation 2020;​142(16_suppl_2):​ N Engl J Med 2013;​369:​2197-206. mothermia after out-of-hospital cardiac
S366-S468. 7. Kirkegaard H, Søreide E, de Haas I, arrest (TTM2): a randomized clinical trial-
3. Hypothermia after Cardiac Arrest et al. Targeted temperature management rationale and design. Am Heart J 2019;​
Study Group. Mild therapeutic hypother- for 48 vs 24 hours and neurologic out- 217:​23-31.
mia to improve the neurologic outcome come after out-of-hospital cardiac arrest: 11. Jakobsen JC, Dankiewicz J, Lange T,
after cardiac arrest. N Engl J Med 2002;​ a randomized clinical trial. JAMA 2017;​ et al. Targeted hypothermia versus target-
346:​549-56. 318:​341-50. ed normothermia after out-of-hospital car-

n engl j med 384;24  nejm.org  June 17, 2021 2293

The New England Journal of Medicine
Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.
 Hypothermia or Normothermia after Cardiac Arrest

diac arrest: a statistical analysis plan. Trials statement from the International Liaison 21. Patterson T, Perkins GD, Hassan Y,
2020;​21:​831. Committee on Resuscitation. Circulation et al. Temporal trends in identification,
12. Iyer VN, Mandrekar JN, Danielson 2018;​137(22):​e783-e801. management, and clinical outcomes after
RD, Zubkov AY, Elmer JL, Wijdicks EFM. 16. van Swieten JC, Koudstaal PJ, Visser out-of-hospital cardiac arrest: insights
Validity of the FOUR score coma scale in MC, Schouten HJ, van Gijn J. Interobserver from the Myocardial Ischaemia National
the medical intensive care unit. Mayo Clin agreement for the assessment of handi- Audit Project database. Circ Cardiovasc
Proc 2009;​84:​694-701. cap in stroke patients. Stroke 1988;​19:​ Interv 2018;​11(6):​e005346.
13. Jacobs I, Nadkarni V, Bahr J, et al. 604-7. 22. Sterne JAC, Savović J, Page MJ, et al.
Cardiac arrest and cardiopulmonary re- 17. Lilja G, Nielsen N, Ullén S, et al. Pro- RoB 2: a revised tool for assessing risk of
suscitation outcome reports: update and tocol for outcome reporting and follow- bias in randomised trials. BMJ 2019;​366:​
simplification of the Utstein templates up in the Targeted Hypothermia versus l4898.
for resuscitation registries: a statement for Targeted Normothermia after Out-of- 23. Button KS, Ioannidis JPA, Mokrysz C,
healthcare professionals from a task force Hospital Cardiac Arrest trial (TTM2). Re- et al. Power failure: why small sample size
of the International Liaison Committee suscitation 2020;​150:​104-12. undermines the reliability of neurosci-
on Resuscitation (American Heart Asso- 18. Jakobsen JC, Gluud C, Wetterslev J, ence. Nat Rev Neurosci 2013;​14:​365-76.
ciation, European Resuscitation Council, Winkel P. When and how should multiple 24. Pereira TV, Horwitz RI, Ioannidis JPA.
Australian Resuscitation Council, New imputation be used for handling missing Empirical evaluation of very large treat-
Zealand Resuscitation Council, Heart and data in randomised clinical trials — ment effects of medical interventions.
Stroke Foundation of Canada, InterAmer- a practical guide with flowcharts. BMC JAMA 2012;​308:​1676-84.
ican Heart Foundation, Resuscitation Med Res Methodol 2017;​17:​162. 25. Olai H, Thornéus G, Watson H, et al.
Councils of Southern Africa). Circulation 19. R Core Team. R: a language and envi- Meta-analysis of targeted temperature
2004;​110:​3385-97. ronment for statistical computing. Vienna:​ management in animal models of car­
14. Gøtzsche PC. Blinding during data Foundation for Statistical Computing, diac arrest. Intensive Care Med Exp 2020;​
analysis and writing of manuscripts. Con- 2013. 8:​3.
trol Clin Trials 1996;​17:​285-290, discus- 20. Cronberg T, Greer DM, Lilja G, Mou- 26. Kelly FE, Nolan JP. The effects of mild
sion 290-293. laert V, Swindell P, Rossetti AO. Brain induced hypothermia on the myocardium:
15. Haywood K, Whitehead L, Nadkarni injury after cardiac arrest: from prognos- a systematic review. Anaesthesia 2010;​65:​
VM, et al. COSCA (Core Outcome Set for tication of comatose patients to rehabili- 505-15.
Cardiac Arrest) in adults: an advisory tation. Lancet Neurol 2020;​19:​611-22. Copyright © 2021 Massachusetts Medical Society.

receive immediate notification when an article

is published online first

To be notified by email when Journal articles

are published online first, sign up at NEJM.org.

2294 n engl j med 384;24  nejm.org  June 17, 2021

The New England Journal of Medicine

Downloaded from nejm.org by STEFFEN STEGHERR on June 16, 2021. For personal use only. No other uses without permission.
Copyright © 2021 Massachusetts Medical Society. All rights reserved.