Hypothermia for neuroprotection in adults after cardiac arrest
Abstract
Background
Good neurological outcome after cardiac arrest is difficult to achieve. Interventions during the resuscitation phase and treatment within the first hours after the event are critical for a favourable prognosis. Experimental evidence suggests that therapeutic hypothermia is beneficial, and several clinical studies on this topic have been published. This review was originally published in 2009; updated versions were published in 2012 and 2016.
Objectives
To evaluate the benefits and harms of therapeutic hypothermia after cardiac arrest in adults compared to standard treatment.
Search methods
We used standard, extensive Cochrane search methods. The latest search date was 30 September 2022.
Selection criteria
We included randomised controlled trials (RCTs) and quasi‐RCTs in adults comparing therapeutic hypothermia after cardiac arrest with standard treatment (control). We included studies with adults cooled by any method, applied within six hours of cardiac arrest, to target body temperatures of 32 °C to 34 °C. Good neurological outcome was defined as no or only minor brain damage allowing people to live an independent life.
Data collection and analysis
We used standard Cochrane methods. Our primary outcome was 1. neurological recovery. Our secondary outcomes were 2. survival to hospital discharge, 3. quality of life, 4. cost‐effectiveness and 5. adverse events. We used GRADE to assess certainty.
Main results
We found 12 studies with 3956 participants reporting the effects of therapeutic hypothermia on neurological outcome or survival. There were some concerns about the quality of all the studies, and two studies had high risk of bias overall. When we compared conventional cooling methods versus any type of standard treatment (including a body temperature of 36 °C), we found that participants in the therapeutic hypothermia group were more likely to reach a favourable neurological outcome (risk ratio (RR) 1.41, 95% confidence interval (CI) 1.12 to 1.76; 11 studies, 3914 participants). The certainty of the evidence was low.
When we compared therapeutic hypothermia with fever prevention or no cooling, we found that participants in the therapeutic hypothermia group were more likely to reach a favourable neurological outcome (RR 1.60, 95% CI 1.15 to 2.23; 8 studies, 2870 participants). The certainty of the evidence was low.
When we compared therapeutic hypothermia methods with temperature management at 36 °C, there was no evidence of a difference between groups (RR 1.78, 95% CI 0.70 to 4.53; 3 studies; 1044 participants). The certainty of the evidence was low.
Across all studies, the incidence of pneumonia, hypokalaemia and severe arrhythmia was increased amongst participants receiving therapeutic hypothermia (pneumonia: RR 1.09, 95% CI 1.00 to 1.18; 4 trials, 3634 participants; hypokalaemia: RR 1.38, 95% CI 1.03 to 1.84; 2 trials, 975 participants; severe arrhythmia: RR 1.40, 95% CI 1.19 to 1.64; 3 trials, 2163 participants). The certainty of the evidence was low (pneumonia, severe arrhythmia) to very low (hypokalaemia). There were no differences in other reported adverse events between groups.
Authors' conclusions
Current evidence suggests that conventional cooling methods to induce therapeutic hypothermia may improve neurological outcomes after cardiac arrest. We obtained available evidence from studies in which the target temperature was 32 °C to 34 °C.
PICOs
Plain language summary
Cooling the body after resuscitation following cardiac arrest
Key message
In this review, we asked whether people resuscitated from cardiac arrest benefit when their bodies are cooled to a temperature of 32 °C to 34 °C. Current evidence suggests that conventional cooling methods to induce hypothermia (low body temperature) may reduce the risk of brain damage and improve neurological outcomes (problems with the nervous system) following successful resuscitation after cardiac arrest.
What is cardiac arrest?
Around 30% to 50% of people with coronary heart disease (when the arteries supplying blood to the heart become narrowed by a build‐up of fatty material within their walls) have sudden cardiac arrest at some stage of their illness. Sudden cardiac arrest means that the heart stops pumping blood and subsequently, circulation of blood throughout the body stops.
How is cardiac arrest treated?
Someone with sudden cardiac arrest needs instant resuscitation to save their life. Resuscitation can be done by people who have no medical training or by healthcare professionals. A variety of techniques can be used, but the first ones are usually to administer chest compressions (pushing hard and frequently on the person's breastbone), rescue breathing techniques (mouth‐to‐mouth resuscitation) and use of a defibrillator that applies electric shocks to the heart to restart it. If people with cardiac arrest are not resuscitated, brain cells begin to be irreversibly damaged, and subsequently, the person dies. After successful resuscitation, treatment within the first few hours is critical to avoid or limit brain damage, and preserve the function and structure of nerve cells in the brain (also called 'neuroprotection'). The symptoms of brain damage vary depending on the severity and duration of the cardiac arrest, as well as the health condition of the person. Symptoms include instant death; coma; paralysis; tremors; difficulty with speech and language; difficulties with thinking, remembering and mental tasks; and body co‐ordination or gait problems. One type of therapy that may help to prevent cell damage consists of cooling the body to 32 °C to 34 °C for several hours after successful resuscitation.
What did we want to find out?
We wanted to know whether people resuscitated from cardiac arrest recover better when their bodies are cooled to a temperature of 32 °C to 34 °C.
What did we do?
We searched medical databases for well‐designed studies that looked at how well people who had their bodies cooled to 32 °C to 34 °C recovered after resuscitation following a cardiac arrest compared to people who were not cooled or not cooled to less than 36 °C.
What did we find?
We included 12 studies (3956 people overall) in our analysis examining the effects of cooling the body after successful resuscitation for cardiac arrest. Eleven studies used conventional cooling methods (for example, cooling pads or ice packs). However, one study used haemofiltration as the cooling method (blood was removed from the body via a tube that passed through a filter to remove toxic substances and a cooling machine before returning to the body – similar to dialysis), so their data could not be summarised with the other studies and were treated separately in the review.
Key results
Cooling the body after successful resuscitation may reduce the risk of brain damage and improve neurological outcomes. When we compared people whose bodies were cooled to 32 °C to 34 °C after resuscitation versus those whose bodies were not cooled, we found that 532 per 1000 of those receiving cooling would have no, or only minor, brain damage, while only 377 per 1000 not receiving cooling would have no, or only minor, brain damage. Cooling had no effect on survival. Cooling the body was associated with an increased risk of pneumonia (an infection that inflames the air sacs in one or both lungs) (384 per 1000 people who had cooling versus 352 per 1000 people who had no cooling), increased risk of low concentrations of blood potassium (185 per 1000 people who had cooling versus 134 per 1000 people who had no cooling) and an increased risk for irregular heartbeats (257 per 1000 people who had cooling versus 184 per 1000 people who had no cooling). Only a few studies looked at these treatable complications.
What are the limitations of the evidence?
Some studies had quality shortcomings including a lack of information on how these studies were carried out and the use of inadequate methods to balance participants between cooling and no cooling groups. However, when we accounted for differences between studies, it became clear that these shortcomings had only a minor impact on the main results, and they did not change the overall findings.
Study funding sources
A dialysis‐related company funded the study that used external cooling. Of the remaining 11 studies included in the main analysis, five received funding from government or non‐profit organisations; two received analysing kits from a company unrelated to cooling and four studies did not provide information on funding.
How up to date is this evidence?
Evidence is current to September 2022.
Authors' conclusions
Summary of findings
| Effects of cooling 32–34 °C on neuroprotection, survival (to discharge or within 6 months) and adverse events for comatose survivors of cardiac arrest | ||||||||||
| Patient or population: adults after cardiopulmonary resuscitation Settings: emergency medicine and intensive care, worldwide Intervention: cooling 32–34 °C (good neurological outcome and survival); cooling via haemofiltration (adverse events) Control: control treatment at ≥ 36 °C | ||||||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect (95% CI) | Number of participants (studies) | Risk of bias | Inconsistency | Indirectness | Imprecision | Other considerations | Overall certainty of evidence | |
|---|---|---|---|---|---|---|---|---|---|---|
| Risk with control | Risk difference | |||||||||
| ≥ 36 °C | Hypothermia at 32–34 °C | |||||||||
| Good neurological outcome | Study population | RR 1.41 | 3914 (11) | Seriousa | Seriousb | Not serious | Not serious | None | ⊕⊕⊖⊖ | |
| 377 per 1000 | 155 more per 1000 | |||||||||
| Survival | Study population | RR 1.07 | 3871 | Seriousa | Seriousb | Not serious | Not serious | None | ⊕⊕⊖⊖ | |
| 448 per 1000 | 31 more per 1000 | |||||||||
| Adverse events – pneumonia | Study population | RR 1.09 | 3634 (4 studies) | Seriousa | Not serious | Not serious | Seriousc | None | ⊕⊕⊖⊖ | |
| 352 per 1000 | 32 more per 1000 | |||||||||
| Adverse events – hypokalaemia | Study population | RR 1.38 | 975 (2 studies) | Seriousa | Not serious | Seriousd | Seriousc | None | ⊕⊖⊖⊖ | |
| 134 per 1000 | 51 more per 1000 | |||||||||
| Adverse events – arrhythmia | Study population | RR 1.40 | 2163 (3 studies) | Seriousa | Not serious | Seriousd | Not serious | None | ⊕⊕⊖⊖ | |
| 184 per 1000 | 73 more per 1000 | |||||||||
| *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and therelative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio. | ||||||||||
| GRADE Working Group grades of evidence | ||||||||||
| aDowngraded one level as there were at least 'some concerns' for all studies (Bernard 2002 and Mori 2000 had high risk of bias). | ||||||||||
Background
Description of the condition
The incidence of out‐of‐hospital sudden cardiac arrest in high‐income countries varies greatly over different study groups and is reported to be between 0.05% and 0.19% per year (Gräsner 2021; McNally 2011). Survival and neurological outcomes are dependent on characteristics of cardiac arrest, the type of emergency medical services (EMS) system, postresuscitation care and the practice of withdrawal of life‐sustaining treatment (WLST). Survival rates from 0% to 18% have been reported (Gräsner 2021). Many people have brain damage due to hypoxic injury. In people with return of spontaneous circulation (ROSC), good functional outcome of cardiac arrest survivors have been reported to be around 11% to 63% (Kim 2022; May 2019; Okubo 2018), and in systems with established WLST, the rate of survival with good neurological outcome was reported to be around 90% (Gräsner 2021). Postresuscitation care aims at optimising therapy and outcomes for cardiac arrest survivors to improve survival, neurological function and quality of life after cardiac arrest including circulation, ventilation, coronary reperfusion, electrolytes, seizure control, temperature management and glucose control as recommended in the main resuscitation guidelines (Nolan 2022; Panchal 2020).
Description of the intervention
Therapeutic hypothermia is also referred to as 'targeted temperature management', 'induced hypothermia', 'temperature control' or 'cooling'. The most recent recommended term is 'temperature control' (Nolan 2021; Nolan 2022; Panchal 2020; Sandroni 2022). As further adaptions of the terminology would mean a title change for our review, we currently stay with the original term 'therapeutic hypothermia' or, more simply, 'cooling'. The intervention is aimed at the preservation of cerebral function in people resuscitated after cardiac arrest. After the person's condition has been stabilised, the body temperature is lowered to 32 °C to 34 °C for 24 hours; other durations have been investigated in randomised controlled trials (RCTs). Conventional cooling comprises surface cooling methods requiring cooling pads, ice packs, water immersion or intravascular cooling with cooling catheters or simply cold fluids. Cooling can be combined with haemofiltration or extracorporeal cardiopulmonary support.
How the intervention might work
Therapeutic hypothermia is believed to work in many ways. Cerebral reperfusion after successful resuscitation is essential and effective in restoring energy stores, but it can trigger harmful chemical cascades, such as the generation of free radicals and other mediators, which leads to multifocal damage to the brain. It was first described by Negovsky as "post‐resuscitation syndrome" (Negovsky 1988). In contrast to accidental hypothermia, therapeutic mild hypothermia (body temperature 32 °C to 34 °C) is administered in a controlled way. Intra‐ischaemic hypothermia for brain protection has been provided for several years with certain surgical procedures and circulatory arrest states. Clinical and experimental results showed protective effects of hypothermia during and after ischaemic situations (Rosomoff 1954). Hypothermia seems to act in a multifactorial way by influencing several damaging pathways simultaneously to reduce cell death within the brain. It mitigates pathophysiological pathways leading to excitotoxicity, apoptosis, inflammation and free radical production, and affects the blood flow, metabolism and blood–brain barrier integrity (Holzer 2010; Yenari 2012). Animal studies on the effects of hypothermia at 32 °C to 36 °C after cardiac arrest showed a favourable effect of postresuscitation hypothermia as compared to normothermia on neurological outcomes with an increasing effect towards a lower target temperature (Arrich 2021).
Why it is important to do this review
The first Cochrane Review was conducted to examine the emergence of mild therapeutic hypothermia as part of routine care for survivors of cardiac arrest (Arrich 2009a). Two RCTs showed that induced hypothermia had a neuroprotective effect in people primarily resuscitated from cardiac arrest (Bernard 2002; HACA 2002), and the International Liaison Committee on Resuscitation (ILCOR) guidelines (Nolan 2003) and other resuscitation guidelines (Deakin 2010; Peberdy 2010) recommended therapeutic hypothermia. After the publication of the targeted temperature management (TTM) trial (Nielsen 2013), the recommended target was changed to 32 °C to 36 °C. After the publication of the TTM2 trial (Dankiewicz 2021), the recommendation was changed to continuous monitoring of core temperature and actively preventing fever (defined as a temperature greater than 37.7 °C) for at least 72 hours (Nolan 2022; Sandroni 2022). The guidelines also concluded that there was insufficient evidence to recommend for or against temperature control at 32 °C to 36 °C or early cooling after cardiac arrest. Therapeutic hypothermia still is a relatively new concept. Studies examining different treatment modalities in different study populations are emerging and are followed by rapid updates of the guidelines; regular independent systematic reviews are increasingly important to reflect the current data and give the best possible basis for treatment recommendations. We present an update of the original Cochrane Review (Arrich 2009a; Arrich 2012; Arrich 2016), which incorporates up‐to‐date evidence.
Objectives
To evaluate the benefits and harms of therapeutic hypothermia after cardiac arrest in adults compared to standard treatment.
Methods
Criteria for considering studies for this review
Types of studies
We included RCTs and quasi‐RCTs. 'Quasi‐randomised' refers to allocation procedures such as alternating days, odd and even days, and the like.
Types of participants
We included adults who experienced a cardiac arrest (regardless of whether in‐hospital or out‐of‐hospital cardiac arrest) and were successfully resuscitated.
We excluded studies of children and adolescents (younger than 18 years), as the presumed pathophysiology of cardiac arrest is typically different from adults.
Types of interventions
The intervention of interest was therapeutic – hypothermia regardless of how body temperature was reduced – applied within six hours of arrival at the hospital. We defined 'therapeutic' as any target body temperature of 32 °C to 34 °C.
We defined the 'control' intervention as standard treatment (at the time of the trial) provided after cardiac arrest.
Types of outcome measures
Outcomes of interest were based on patient‐centred outcomes, availability of data, and the recommendations on Core Outcome Set for Cardiac Arrest (COSCA) in Adults: an Advisory Statement from the International Liaison Committee on Resuscitation (Haywood 2018).
Primary outcomes
-
Neurological recovery. We expected the ideal time point of outcome to be reported as the best neurological outcome during the hospital stay and in cerebral performance categories (CPCs) (Stiell 2009). CPC categories are defined as follows.
-
CPC 1: good cerebral performance: conscious, alert and capable of normal life. Normal cerebral function. May have minor psychological or neurological deficits that do not significantly compromise cerebral or physical function.
-
CPC 2: moderate cerebral disability: conscious, alert, sufficient cerebral function for activities of daily life (e.g. able to dress, travel by public transportation, food preparation). May have hemiplegia, seizures, ataxia, dysarthria, dysphasia, or permanent memory or mental changes.
-
CPC 3: severe cerebral disability: conscious, with at least limited cognition. Dependent on others for daily life support (i.e. institutionalised or at home with exceptional family effort) because of impaired brain function. Includes individuals with a wide range of cerebral abnormalities, from ambulatory participants with severe memory disturbance or dementia precluding independent existence to paralysed participants who can communicate only with their eyes (e.g. the locked‐in syndrome).
-
CPC 4: coma or vegetative state: not conscious, unaware of surroundings, no cognition. No verbal or psychological interaction with the environment. May appear awake because of spontaneous eye‐opening or sleep–wake cycle. Includes individuals showing all degrees of unresponsiveness that are neither CPC three (conscious) nor CPC five (coma, which satisfies brain death criteria).
-
CPC 5: certified brain death.
-
If the study authors grouped the primary outcome into 1 or 2 (good recovery) and 3 to 5 (unfavourable recovery), we adapted it for our meta‐analysis. If the primary outcome was not reported in CPC categories, we accepted reports of 'good' neurological outcomes, and we assumed that this was comparable with a CPC score of 1 or 2.
Secondary outcomes
-
Survival to hospital discharge, at three months, six months and long term
-
Quality of life at six months and long term
-
Cost‐effectiveness
We defined 'long term' as a minimum of one year.
-
Adverse events
We aimed to report adverse events as described by the study authors.
Search methods for identification of studies
Electronic searches
We searched for studies as described in the Cochrane Handbook of Systematic Reviews of Interventions (Lefebvre 2021). We applied no language or publication status restrictions, and we ran the update‐search from January 2015 to 30 September 2022.
For this update we searched the following databases for eligible trials.
-
Cochrane Central Register of Controlled Trials (CENTRAL, 2022, Issue 9; Appendix 1).
-
MEDLINE ALL (OvidSP, 2015 to 30 September 2022; Appendix 2).
-
Excerpta Medica database (Embase) (OvidSP, 2015 to 30 September 2022; Appendix 3).
-
Cumulative Index to Nursing and Allied Health Literature (CINAHL, EBSCO 2015 to 30 September 2022; Appendix 4).
-
Web of Science and Biosis Previews (Clarivate, 2015 to 30 September 2022; Appendix 5).
We adapted the original MEDLINE search strategy to the rest of the databases. In MEDLINE, Embase and CINAHL we combined the searches with RCT filters recommended by the Cochrane Handbook of Systematic Reviews of Interventions (Lefebvre 2021). The original searches were performed in January 2007 (Arrich 2009a), updated in July 2011 (Arrich 2012) and in May 2015 (Arrich 2016).
Searching other resources
We searched for relevant systematic reviews on the topic in Epistemonikos (Appendix 6).
We searched the bibliographic references and citations of included studies and systematic reviews for other potentially eligible studies. We used PubMed and the universities' library services for the citation searches.
We searched two trial registers for unpublished and ongoing studies.
-
ClinicalTrials.gov (www.clinicaltrials.gov/; Appendix 7).
-
World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (trialsearch.who.int/; Appendix 8).
We contacted trial authors and experts in the field to ask for information on ongoing, unpublished or published trials on the topic. We searched for errata and retractions for included studies in PubMed and Retraction Watch Database (retractiondatabase.org/). The Information Specialist from the Cochrane Emergency and Critical Care Group update the search strategy, which the Information Specialist from the Cochrane Haematology Group peer reviewed.
Data collection and analysis
Selection of studies
We used Covidence for the deduplication of the references before screening the search result (www.covidence.org/). At least two review authors (from JA, JO, NS, and MH) independently scanned each reference for inclusion in the review. One review author (HH) acted as arbiter in case of discrepancies.
Data extraction and management
We extracted data using a data extraction form (see Appendix 9). Four review authors (JA, NS, JO, MH) independently entered all relevant data into a data extraction form and compared entries. One review author transferred the data into Review Manager Web, with the other review authors double checking the entries (RevMan Web 2022). We resolved any disagreements by discussion.
We entered the following variables into Review Manager Web (RevMan Web 2022).
-
Study and participant characteristics
-
Data on outcomes
-
Items on risk of bias
-
Within‐study subgroup effect estimates
Where a review author was also a contributor to an included study, that review author was not involved in the data extraction and management process.
Assessment of risk of bias in included studies
We used the Cochrane RoB 2 tool (Sterne 2019), and guidance of the Cochrane Handbook of Systematic Reviews of Interventions (Higgins 2022), to determine risk of bias.
Measures of treatment effect
We calculated risk ratios (RR) and their 95% confidence intervals (CIs).
Unit of analysis issues
We excluded cluster‐randomised trials from the analyses. In the case of multiple treatment groups, we planned to combine groups to create a single comparison; by nature of the condition and outcomes, we identified no cross‐over trials.
When feasible, we planned to transform scales of outcome data and add to the analysis to provide the most comprehensive picture of the evidence.
Dealing with missing data
We used the intention‐to‐treat (ITT) principle for all analyses. If data were missing, we attempted to obtain the information from the study authors. We included assessment of loss to follow‐up in our quality assessment and reported this in the Characteristics of included studies table. If there was a considerable quantity of data missing, we planned to investigate the possible mechanism of the missing data (whether random or not). We planned to perform a sensitivity analysis to assess the influence of this possible selection bias on our estimates.
Assessment of heterogeneity
We assessed data for clinical and statistical heterogeneity. Clinical heterogeneity may be caused by differences in study populations, interventions and controls, or by variable definitions of the endpoint (Thompson 1995). Pooling of data may not be suitable in cases of severe heterogeneity.
Assessment of reporting biases
We assessed the presence of possible publication bias and heterogeneity using funnel plots (plotting effect versus precision) (Egger 1997), and planned to present this information if we identified more than 10 trials per analysis.
Data synthesis
We planned to perform a quantitative data synthesis using standard statistical procedures provided in Review Manager Web (RevMan Web 2022). The principal summary effect estimate was the RR with a 95% CI.
In the case of negligible heterogeneity, we used a fixed‐effect model to calculate summary effects; otherwise, we used random‐effects models. We assessed statistical heterogeneity using the I2 statistic (Higgins 2003), and considered statistical heterogeneity relevant with an I2 statistic greater than 50%.
Subgroup analysis and investigation of heterogeneity
For the primary endpoint (neurological recovery) and survival, we performed the following subgroup analyses.
-
Treatment of the control group: hypothermia versus no temperature management or fever prevention
-
Cause of cardiac arrest (presumed cardiac versus non‐cardiac)
-
Location of arrest (in‐hospital versus out‐of‐hospital)
-
Primary electrocardiography (ECG) rhythm (ventricular fibrillation (VF) versus other)
-
Witnessed versus non‐witnessed arrest
-
Bystander cardiopulmonary resuscitation (CPR) rate (70% or greater, 43% to 59% and 7% to 42%)
-
No‐flow time (time from cardiac arrest to start of chest compressions) (up to one minute, one to two minutes, and three to five minutes)
-
Duration of hypothermia (four hours, 12 to 28 hours and 72 hours)
-
Time interval from ROSC to intervention (intervention started within two hours of ROSC and time interval from ROSC to intervention not reported)
Sensitivity analysis
We performed sensitivity analyses to examine the influence of the overall study quality on the effect estimate of the main analysis without consideration of heterogeneity and publication status. To investigate whether the model choice might influence our results, we compared estimates derived from random‐effects models versus those obtained from fixed‐effect models.
Summary of findings and assessment of the certainty of the evidence
We used GRADEpro GDT to interpret our findings (GRADEpro GDT; Langendam 2013) and to create summary of findings Table 1. This table provides outcome‐specific information concerning the overall certainty of evidence gathered from studies included in the comparisons, the magnitude of effects of the interventions examined and the sum of available data on important outcomes.
We included the following outcomes in the summary of findings table: neurological outcome, survival (conventional cooling versus no cooling) and adverse events (all studies).
Results
Description of studies
Results of the search
Our updated search (to September 2022) resulted in 3689 additional hits (duplicates excluded) (see Figure 1). From these, we excluded 3659 records according to our eligibility criteria by assessing the abstract or the title. Thirty records remained for full‐text assessment and we excluded 25 articles with reasons (see Characteristics of excluded studies table). One full‐text article contained results from two studies (Hachimi‐Idrissi 2005 LSP; Hachimi‐Idrissi 2005 SSP). We added six new studies (five reports) from the updated search to the six studies from the previous version of the review.
Study flow diagram.
The quantitative analysis included 12 RCTs and quasi‐RCTs with 3956 participants (Bernard 2002; Dankiewicz 2021; HACA 2002; Hachimi‐Idrissi 2001; Hachimi‐Idrissi 2005 LSP; Hachimi‐Idrissi 2005 SSP; Kwon 2021; Lascarrou 2019; Laurent 2005; Mori 2000; Nielsen 2013; Zhang 2005) (see Characteristics of included studies table).
The decision on the inclusion of studies with an active control group treatment was not unequivocal as active temperature management at 36 °C is different from 'no temperature management 'or fever control (Kwon 2021; Mori 2000; Nielsen 2013). It may be debatable whether temperature control at 36 °C can be seen as 'standard therapy' as we outlined in our inclusion criteria for the control group (see also Discussion). We decided to include these studies as subgroup analyses to provide a comprehensive picture of the available evidence.
Included studies
We included 12 studies in this updated review (Bernard 2002; Dankiewicz 2021; HACA 2002; Hachimi‐Idrissi 2001; Hachimi‐Idrissi 2005 LSP; Hachimi‐Idrissi 2005 SSP; Kwon 2021; Lascarrou 2019Laurent 2005; Mori 2000; Nielsen 2013; Zhang 2005) (see Characteristics of included studies table)
Bernard 2002 was a quasi‐RCT carried out by the ambulance service of Melbourne (Australia) and four adjacent emergency departments and intensive care units (ICUs) over three years. It included 77 participants with out‐of‐hospital cardiac arrest and VF as their first cardiac rhythm who were comatose after they achieved ROSC. The mean age of participants was 66 years and 33% were women. Ambulance staff randomly assigned participants to receive 1. therapeutic hypothermia using ice packs placed around the head, neck, torso and limbs or 2. standard therapy without hypothermia. The target temperature of the intervention group was 33 °C over 12 hours; this was followed by a passive rewarming phase over eight hours. Outcome parameters were survival with good neurological function to hospital discharge; in‐hospital mortality; haemodynamic, biochemical and haematological effects of hypothermia; and adverse events to hospital discharge.
Dankiewicz 2021 was an RCT undertaken in 61 university and community hospitals in Australia, Europe and the USA. It included 1900 comatose survivors of out‐of‐hospital cardiac arrest of presumed cardiac or unknown cause. Participants were randomised to 1. targeted hypothermia at 33 °C, followed by controlled rewarming, or 2. targeted normothermia with early treatment of fever (body temperature 37.8 °C or greater). TTM was applied by intravenous cold (4 °C) fluids, endovascular cooling devices with closed‐loop systems and surface cooling devices with a closed loop. TTM was maintained for 28 hours after randomisation. The outcomes were death and neurological function at six months.
HACA 2002 was an RCT carried out by nine centres in five European countries and four adjacent emergency departments and ICUs over five years. It included 275 participants with out‐of‐hospital bystander‐witnessed cardiac arrest of presumed cardiac cause, VF or non‐perfusing ventricular tachycardia (VT) as first cardiac rhythm, who were comatose after resuscitation. The mean age of participants was 59 years and 24% were women. After admission, participants were randomly assigned to receive 1. therapeutic hypothermia by an external cooling device or 2. standard therapy without hypothermia. The target temperature of the intervention group was 32 °C to 34 °C maintained over 24 hours; this was followed by a passive rewarming phase over eight hours. The outcome parameters were survival, neurological outcome and adverse events.
Hachimi‐Idrissi 2001 was a feasibility trial carried out at a 700‐bed teaching hospital in Brussels (Belgium) for six months. It included 33 consecutive participants with out‐of‐hospital cardiac arrest and asystole or pulseless electrical activity (PEA) of presumed cardiac origin who achieved a ROSC. The mean age of participants was 72 years and 39% were women. After admission and stabilisation, participants were randomly assigned to 1. a helmet device placed around the head and neck containing a solution of aqueous glycerol or 2. standard therapy without hypothermia. The target temperature of the intervention group was 34 °C, maintained for four hours; this was followed by a passive rewarming phase over eight hours. Outcome parameters were laboratory values and haemodynamics, survival to hospital discharge, overall performance categories (OPCs) and adverse events. The authors provided the completed data of all 33 participants; the publication reported on only 30 participants, as the follow‐up was not completed at the time of submission.
Hachimi‐Idrissi 2005 LSP and Hachimi‐Idrissi 2005 SSP were two studies published in one report. There were two distinct study populations, each randomised individually, receiving different interventions. It was carried out at the Department of Emergency Medicine of the Vrije Universiteit Brussels (Belgium). In the short study period (SSP) study, 33 adult surviving asystole or PEA were randomised to 1. hypothermia of 33 °C for approximately four hours or 2. normothermia (Hachimi‐Idrissi 2005 SSP). Hypothermia was applied by a helmet device placed on the participant's head. In the long study period (LSP) study, 28 adults with witnessed cardiac arrest of cardiac origin and VF or non‐perfusing VT were randomised to 1. hypothermia at 33 °C for 24 hours or 2. normothermia at 37 °C (Hachimi‐Idrissi 2005 LSP). Hypothermia was applied by a mattress with a cover that delivered cool air over the entire body. The main outcome parameters were neurological outcome and survival at six months.
Kwon 2021 was an RCT carried out in two referral hospitals in Seoul (South Korea). Fifty‐seven adults surviving cardiac arrest from presumed cardiac causes were randomised to TTM at either 1. 33 °C or 2. 36 °C and applied within one hour after ROSC by an automated external cooling device. Participants were maintained at the target temperature for 24 hours and were then rewarmed at a rate of 0.25 °C/hour to 37.5 °C. The temperature of the control group treatment was 36 °C. Outcomes were survival and neurological outcomes at six months.
Lascarrou 2019 was an RCT carried out in 25 ICUs in France. It included 584 comatose survivors of in‐ and out‐of‐hospital cardiac arrest with non‐shockable rhythm due to any cause. Participants were randomised to receive 1. therapeutic hypothermia at 33 °C maintained for 24 hours or 2. control. Cooling methods included active internal or external cooling using a specific device, and active external cooling without a specific device. Infusion of cold fluids (4 °C) was recommended to expedite the achievement of the target temperature. The rewarming rate was 0.25 °C to 0.5 °C/hour, 37 °C was then maintained for an additional 24 hours. The participants in the control group were maintained at 36.5 °C to 37.5 °C for 48 hours. The outcomes were survival and neurological outcome at three months.
Laurent 2005 was an RCT carried out at two ICUs in Paris (France) over two years. It included 42 comatose survivors of out‐of‐hospital cardiac arrest of presumed cardiac cause, VF or asystole as first cardiac rhythm. The mean age of participants was 54 years and 19% were women. After admission, participants were randomly assigned to receive 1. high‐flow haemofiltration, 2. high‐flow haemofiltration plus therapeutic hypothermia or 3. standard therapy without hypothermia. The target temperature of the intervention group was 32 °C to 33 °C maintained over 24 hours. Outcome parameters were survival, neurological outcome and adverse events.
Mori 2000 was published as an abstract with additional information provided by study authors, Kazuhisa Mori and Eric Dickson. The study was carried out at a University hospital in Sapporo (Japan). It included 54 participants with out‐of‐hospital cardiac arrest with a Glasgow Coma Scale (GCS) score of less than 8 after resuscitation. After admission, participants were randomly assigned to receive 1. therapeutic hypothermia at 32 °C to 34 °C by water‐circulating blankets and an ice‐mounted blanket over the participant or 2. standard therapy at 36 °C. The intervention was maintained over three days. The reported outcome parameter was neurological outcome.
Nielsen 2013 was carried out at 36 ICUs in Europe and Australia over three years. It included 950 adults with out‐of‐hospital cardiac arrest of presumed cardiac cause and sustained ROSC who were comatose after resuscitation (GCS less than 8). The mean age of participants was 64 years and 19% were women. After admission, participants were randomly assigned to receive 1. temperature management at 33 °C (method of choice was up to the discretion of treating physicians) or 2. temperature management at 36 °C (method of choice was up to the discretion of treating physicians). The target temperature of the intervention group was 32 °C to 34 °C over 24 hours; this was followed by a rewarming phase over eight hours. Outcome parameters were survival, and neurological and functional outcomes. Cronberg 2015b was a substudy that provided an exploratory analysis of the cognitive function and quality of life of the participants included in the TTM trial (Nielsen 2013). It was included as a report of Nielsen 2013.
Zhang 2005 was a small RCT including 16 resuscitated adults randomised to 1. hypothermia and 2. normothermia. The hypothermia group were cooled to 33 °C for 72 hours and rewarmed by 0.125 °C/hour. Hypothermia was applied by administering chlorpropham, an electronic ice blanket and an ice cap. Participants in the normothermia group were treated at room temperature. Data on neurological outcomes were presented as the Barthel Index in a continuous format. Assuming an approximate normal distribution for the Barthel Index, we calculated the probability of being above or below the cut‐off of 60 given the mean and standard deviation (SD) from the report. We used the standard formula z = (x0 − x)/SD, which we compared to the areas in one tail (z‐>P) of the normal distribution (Altman 1991).
Clinical heterogeneity
We identified clinical heterogeneity due to cooling methods.
-
In contrast to the other studies, Laurent 2005 used haemofiltration as the mode of cooling, which is different from the standard cooling methods used in the other RCTs; therefore, we did not pool data with those from the remaining studies.
-
We included studies with active control group management at 36 °C as subgroups of the main analysis (Kwon 2021; Mori 2000; Nielsen 2013).
Based on cooling methods and clinical characteristics, the remaining studies appeared appropriate for pooling in the main analysis (Bernard 2002; Dankiewicz 2021; HACA 2002; Hachimi‐Idrissi 2001; Hachimi‐Idrissi 2005 LSP; Hachimi‐Idrissi 2005 SSP; Kwon 2021; Lascarrou 2019; Mori 2000; Nielsen 2013; Zhang 2005).
Excluded studies
We excluded 25 studies; most were duplicates, the intervention did not apply to our criteria (e.g. prehospital cooling) or the study was not an RCT. See the reasons for exclusion in the Characteristics of excluded studies table.
Awaiting classification
Ongoing studies
We identified no ongoing studies.
Risk of bias in included studies
We assessed all trials using the Cochrane RoB 2 tool (Sterne 2019). It evaluates possible bias for five items: randomisation process, missing outcome data, deviations from intended interventions, measurement of the outcome and selection of the reported results (see Figure 2; Risk of bias table for Analysis 1.1; Risk of bias table for Analysis 1.2; Risk of bias table for Analysis 2.1; Risk of bias table for Analysis 2.2).
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Mori 2000 was an abstract that provided only scarce information on essential quality criteria. Bernard 2002 showed deficits in the randomisation process and made adjustments for inequalities in baseline characteristics between treatment and control groups. These two studies had an overall assessment of high risk of bias. All other studies were assessed to have some concerns, which were mainly due to the inability to blind carers and people delivering the interventions and the possible implications for the delivery of the intervention, and the lack of registration of protocols, particularly in older studies.
Assessment of reporting bias
Given the high degree of heterogeneity and the small number of studies in the individual analyses we did not present funnel plots.
Effects of interventions
Primary outcome
Good neurological outcome
Overall eight studies (2870 participants) reported on conventional cooling methods compared with no cooling or fever prevention (Bernard 2002; Dankiewicz 2021; HACA 2002; Hachimi‐Idrissi 2001; Hachimi‐Idrissi 2005 LSP; Hachimi‐Idrissi 2005 SSP; Lascarrou 2019; Zhang 2005). The pooled analysis showed a better neurological outcome for the conventional cooling group (RR 1.60, 95% CI 1.15 to 2.23; I2 = 68%; Figure 3). The certainty of the evidence was low. Most of the heterogeneity was caused by the inclusion of Dankiewicz 2021.
Forest plot: conventional cooling versus control: 1.1 neurological outcome.
Three studies evaluating the effects of conventional cooling versus TTM at 36 °C found no evidence of a difference in neurological outcome (RR 1.78, 95% CI 0.70 to 4.53; I2 = 73%; Figure 3) (Kwon 2021; Mori 2000; Nielsen 2013). The certainty of the evidence was low.
We pooled data from 11 studies that used conventional cooling methods regardless of the control treatment (Bernard 2002; Dankiewicz 2021; HACA 2002; Hachimi‐Idrissi 2001; Hachimi‐Idrissi 2005 LSP; Hachimi‐Idrissi 2005 SSP; Kwon 2021; Lascarrou 2019; Mori 2000; Nielsen 2013; Zhang 2005). The analysis found that cooling to 33 °C was superior to no cooling, fever control or temperature management at 36 °C (RR 1.41, 95% CI 1.12 to 1.76; Figure 3). The certainty of the evidence was low. We found considerable heterogeneity when pooling these studies (I² = 67%), which was mainly driven by the inclusion of the TTM1 trial (Nielsen 2013), and TTM2 trial (Dankiewicz 2021).
For this outcome there were some concerns of bias in most studies, two studies had a high risk of bias (Bernard 2002; Mori 2000). We downgraded the certainty of the evidence for this outcome because of the risk of bias and inconsistency, which resulted in an overall low certainty of evidence (summary of findings Table 1).
One study included participants undergoing haemofiltration after cardiac arrest (Laurent 2005). This study compared three treatment groups: 1. high‐flow haemofiltration plus normothermia, 2. high‐flow haemofiltration plus therapeutic hypothermia and 3. standard therapy without hypothermia. We compared high‐flow haemofiltration versus high‐flow haemofiltration plus therapeutic hypothermia for the analysis and excluded standard therapy without hypothermia. As the intervention was distinctively different from the cooling methods in all other studies it introduced considerable clinical heterogeneity which prevented pooling with the studies described above. There was no evidence of a difference in good neurological outcome between groups (RR 0.71, 95% CI 0.32 to 1.54; Figure 4). For this study, there were some concerns due to the risk of bias.
Forest plot: haemofiltration cooling versus haemofiltration normothermia: 1.2: neurological outcome.
Secondary outcomes
Survival to hospital discharge, at three months, six months and long term
Seven studies (2875 participants) reported on the effects of conventional cooling methods compared to control (no cooling or fever prevention) (Bernard 2002; Dankiewicz 2021; HACA 2002; Hachimi‐Idrissi 2001; Hachimi‐Idrissi 2005 LSP; Hachimi‐Idrissi 2005 SSP; Lascarrou 2019). The pooled result showed no survival benefit for the conventional cooling group (RR 1.17, 95% CI 0.96 to 1.42; I2 = 48%; Figure 5). The certainty of the evidence was low. Statistical heterogeneity was driven by one study (Dankiewicz 2021).
Forest plot: conventional cooling versus control: 1.2: survival.
Two studies (996 participants) reported the effects of conventional cooling compared with TTM at 36 °C (Kwon 2021; Nielsen 2013). The pooled result showed no survival benefit for the conventional cooling group (RR 0.98, 95% CI 0.86 to 1.10; I2 = 0%; Figure 5). The certainty of the evidence was low.
There were some concerns due to the risk of bias for most studies with this outcome, and one had high risk of bias (Bernard 2002).
We downgraded the certainty of the evidence for this outcome because of the risk of bias and inconsistency, which resulted in an overall low certainty of evidence (summary of findings Table 1).
There was no evidence of a difference in survival between cooling using haemofiltration and haemofiltration with normothermia (RR 0.71, 95% CI 0.32 to 1.54; Figure 6). For this study, there were some concerns about the risk of bias (Risk of bias table for Analysis 2.2).
Forest plot: haemofiltration cooling versus haemofiltration normothermia: 2.2 survival.
When we pooled data from all studies that used conventional cooling methods regardless of the control treatment, there was no survival benefit for cooling at 33 °C (RR 1.07, 95% CI 0.95 to 1.20; P = 0.26; Figure 5) (Bernard 2002; Dankiewicz 2021; HACA 2002; Hachimi‐Idrissi 2001; Hachimi‐Idrissi 2005 LSPHachimi‐Idrissi 2005 SSP; Kwon 2021; Lascarrou 2019; Nielsen 2013). Heterogeneity was moderate (I² = 34%). The certainty of the evidence was low.
We found no data long‐term survival.
Quality of life at six months and long term
Two studies reported quality of life at six months. The outcomes were in part reported as a summary result of one scale and as separate components of another scale. Cronberg 2015b was a substudy to Nielsen 2013, and compared TTM at 33 °C versus TTM at 36 °C. The study reported two items of the 36‐item Short Form‐36 (SF‐36) Version 2 Mental and Physical Component Summary. The mean Mental Component Summary score of the SF‐36v2 was 49.1 (SD 12.5) for survivors in the 33 °C group compared with 49.0 (SD 12.2) in the 36 °C group (P = 0.77). The mean Physical Component Summary scores were 46.8 (SD 13.8) for survivors in the 33 °C group and 47.5 (SD 13.8) in the 36 °C group (P = 0.44). Dankiewicz 2021 reported that there was no difference between the groups when assessing quality of life using two scoring systems: 1. including participants who died during the observation period (with the score on the five level EQ‐5D (EQ‐5D‐5L) Visual Analogue Scale set to 0); 2. only those who survived (mean between‐group difference in participants who survived to six months −0.8 points, 95% CI −3.6 to 2.0).
Cost‐effectiveness
We found no data on cost‐effectiveness.
Adverse events
We included all trials that reported adverse events, regardless of heterogeneity. Seven studies (3788 participants) reported on 26 different groups of adverse events (Table 1; summary of findings Table 1) (Bernard 2002; Dankiewicz 2021; HACA 2002; Hachimi‐Idrissi 2001; Lascarrou 2019; Laurent 2005; Nielsen 2013). For most adverse events, there was no evidence of differences between groups.
| Outcome or subgroup | Studies | Participants | Risk ratio (95% CI) |
|---|---|---|---|
| Bleeding of any severity | 4 | 3636 | 1.09 (0.94 to 1.27) |
| Need for platelet transfusion | 1 | 273 | 5.11 (0.25 to 105.47) |
| Significant haemorrhagic complications | 1 | 77 | Not estimable |
| Pneumonia | 4 | 3634 | 1.09 (1.00 to 1.18) |
| Pancreatitis | 1 | 273 | 0.51 (0.05 to 5.57) |
| Sepsis | 3 | 3054 | 1.17 (0.94 to 1.45) |
| Septic shock | 1 | 933 | 0.87 (0.50 to 1.52) |
| Renal failure or oliguria | 2 | 303 | 0.88 (0.48 to 1.61) |
| Haemodialysis | 4 | 1869 | 1.12 (0.85 to 1.48) |
| Seizures | 3 | 1783 | 1.11 (0.95 to 1.30) |
| Severe, haemodynamic compromising or long‐lasting arrhythmia | 3 | 2163 | 1.40 (1.19 to 1.64) |
| Any arrhythmia | 1 | 933 | 0.98 (0.93 to 1.04) |
| Pulmonary oedema | 2 | 850 | 0.93 (0.57 to 1.52) |
| Cardiac complications | 1 | — | No totals |
| Hypokalaemia | 2 | 975 | 1.38 (1.03 to 1.84) |
| Hypophosphataemia | 2 | 975 | 1.10 (0.92 to 1.33) |
| Hypoglycaemia | 1 | 933 | 1.12 (0.64 to 1.97) |
| Hypomagnesaemia | 1 | 933 | 1.20 (0.88 to 1.65) |
| Pressure sores | 1 | 269 | Not estimable |
| Skin complications related to device | 1 | 1849 | 1.99 (0.68 to 5.80) |
| Bacteraemia | 1 | 581 | 1.14 (0.51 to 2.54) |
| Central venous catheter infection | 1 | 581 | 2.09 (0.39 to 11.33) |
| Urinary tract infections | 1 | 581 | 0.78 (0.34 to 1.83) |
| Nosocomial infections, other than central venous catheter infection and urinary tract infections | 1 | 581 | 1.05 (0.31 to 3.57) |
| Ventilator‐associated pneumonia | 1 | 581 | 1.31 (0.87 to 1.99) |
| Vasopressors between day 0 and 7 | 1 | 581 | 1.01 (0.94 to 1.09) |
CI: confidence interval.
Four studies (3634 participants) reported pneumonia, which may have been more frequent in the hypothermia group (RR 1.09, 95% CI 1.00 to 1.18) (Dankiewicz 2021; HACA 2002; Lascarrou 2019; Nielsen 2013). The certainty of the evidence was low.
Three studies (2163 participants) reported severe, haemodynamically compromising or long‐lasting arrhythmia, which was more frequent in the hypothermia group (RR 1.40, 95% CI 1.19 to 1.64) (Dankiewicz 2021; HACA 2002; Laurent 2005). The certainty of the evidence was low.
Two studies (975 participants) reported hypokalaemia, which was more frequent in the hypothermia group (RR 1.38, 95% C 1.03 to 1.84) (Laurent 2005; Nielsen 2013). The certainty of the evidence was very low.
Subgroup analyses of condition and intervention
The results of the subgroup analyses are listed in Analysis 3.1; Analysis 3.2; Analysis 3.3; Analysis 3.4; Analysis 3.5; Analysis 3.6 (it should be noted that in the table the summary number of studies in each subgroup refers to the number of studies, not the number of comparisons). We considered all studies providing results on the primary outcome and additional available data from the study authors. The results comprised fewer studies per group and a more selected view of the available data and need to be interpreted with care. The effect of hypothermia seemed to be highest in populations with non‐witnessed cardiac arrest (Analysis 3.4), bystander CPR rates of less than 60% (Analysis 3.5), no‐flow times of more than one minute (Analysis 3.6), and when hypothermia was initiated within two hours after ROSC (Analysis 3.8).
Sensitivity analyses
The results from the sensitivity analyses can be found in Table 2.
| Sensitivity analysis | Measure | Risk ratio (95% CI) |
|---|---|---|
| Robustness against risk of bias in individual studies | Excluding studies with high risk of bias (Bernard 2002; Mori 2000) | 1.29 (1.04 to 1.60) |
| Robustness against methods of the intervention | Pooling with studies using haemofiltration (Laurent 2005) | 1.35 (1.09 to 1.67) |
| Robustness against model choice | Using fixed‐effect model instead of random‐effects model | 1.09 (1.02 to 1.18) |
| Robustness against including converted data | Excluding converted study data from a continuous outcome (Zhang 2005) | 1.33 (1.07 to 1.65) |
CI: confidence interval.
Robustness against the risk of bias in individual studies
Two studies (113 participants) had an overall high risk of bias. One study had inadequate allocation concealment (Bernard 2002), and one study had missing information on allocation concealment (Mori 2000; 54 participants). Excluding these studies from the main analysis reduced the effect size (from RR 1.60 to RR 1.40, 95% CI 1.14 to 2.25 for studies with control group treatment above 36 °C, from RR 1.41 to RR 1.29, 95% CI 1.04 to 1.60 for all studies). The overall effects appeared robust against possible deficits in allocation concealment.
Robustness against methods of the intervention
Laurent 2005 included participants undergoing haemofiltration after cardiac arrest, a method that was relevantly different from all other studies using conventional cooling. This prevented the pooling of their data with the rest of the study data. In the sensitivity analysis, pooling did not result in a relevant change of the summary effect (from RR 1.41 to 1.35, 95% CI 1.09 to 1.67).
Robustness against model choice
We used random‐effects models because of considerable heterogeneity (I² = 67%). Using fixed‐effect models would have been inadequate and resulted in a reduced effect size (RR 1.09, 95% CI 1.02 to 1.18).
Robustness against including converted data
Zhang 2005 was a small study presenting data on neurological outcomes in a continuous format. Their data were converted to odds ratios, so we could add their data to the analysis. Excluding their data resulted in a reduction of the effect size from RR 1.41 to 1.33, without affecting the 95% CIs.
Discussion
Summary of main results
Low‐certainty evidence from 12 studies comprising 3956 participants contributing data to the primary outcome of this review suggests that therapeutic hypothermia with conventional cooling methods may improve neurological outcome after cardiac arrest.
Overall completeness and applicability of evidence
All 12 studies reported on the primary outcome (Bernard 2002; Dankiewicz 2021; HACA 2002; Hachimi‐Idrissi 2001; Hachimi‐Idrissi 2005 LSP; Hachimi‐Idrissi 2005 SSP; Kwon 2021; Lascarrou 2019; Laurent 2005; Mori 2000; Nielsen 2013; Zhang 2005).
One of the challenges of this review was heterogeneity at the study level. In the case of Laurent 2005, the two treatment modalities (haemodialysis with and without mild therapeutic hypothermia) were clinically too heterogeneous to be combined with those of other studies using conventional cooling. As mentioned in the Background section, one of the theories of the beneficial effects of cooling deals with the attenuation of the effects of free radicals and other mediators. Haemofiltration may act similarly by reducing the number of free radicals. This effect could interact with the effect of therapeutic hypothermia. RCTs have shown that haemofiltration, especially at higher volumes, is successful at treating hypercytokinaemia in critically ill people, but the effect on clinical outcomes is undetermined (Lou 2018).
Control group treatment was another important driver for heterogeneity. Kwon 2021, Mori 2000, and Nielsen 2013 differed in terms of control group treatment, as investigators used active TTM at 36 °C in contrast to no temperature management or fever control in the other studies. Despite seemingly subtle differences, this can be expressed as two distinct clinical questions: 'Does mild therapeutic hypothermia compared with no mild therapeutic hypothermia have any influence on outcomes?' versus 'Does temperature management targeted to 33 °C compared with 36 °C influence outcomes?' These questions were often ignored in clinical discussions. To deal with concerns of heterogeneity and, at the same time, give a better picture of available data, we decided to include all studies and perform subgroup analyses, as well as an overall analysis summarising all relevant evidence. Without temperature management in the control group, therapeutic hypothermia at 32 °C to 34 °C improves neurological outcome after cardiac arrest, and this effect persists when the control group receives active temperature management at 36 °C. Whether 36 °C, cooling to 32 °C to 34 °C and no cooling are equally effective can be judged only by an indirect comparison. However, this indirect comparison should be performed when a sufficient degree of transitivity is present, requiring that different sets of RCTs are similar, on average, in "all important factors other than the intervention comparison being made" (Chaimani 2022) (see also Agreements and disagreements with other studies or reviews; Storm 2014; Stub 2014).
In our sensitivity analysis, the effect was attenuated by allocation concealment.
Wolfrum 2022 was released after our review was submitted for the editorial process. We have therefore designated the study as "awaiting classification" and it will be included in the next update. Pending formal assessment it seems that including its result would not have changed the main conclusion.
Quality of the evidence
There was at least 'some concern' in the overall risk of bias assessment for all studies, and two studies had an overall high risk of bias. In the analysis including studies with control group treatment above 36 °C as well as in the analysis including all studies, the overall effects appeared robust against possible risks of bias.
In comparison to the last version of this review, the certainty of the evidence and inconsistency were the main reasons for downgrading the evidence. We reassessed the risk of bias using the RoB 2 tool, which resulted in a more granular view of the available studies and a heavier weight on grading.
Inconsistency was serious and caused a downgrading of the certainty of the evidence for the outcomes of neurological function and survival. Two studies were the main drivers for heterogeneity (Dankiewicz 2021; Nielsen 2013). Adding them to the analysis led to an increase in statistical heterogeneity from 2% to 68% in Analysis 1.1.1 and 0% to 73% in Analysis 1.1.2. Inconsistency was less pronounced for adverse events and in the subgroup analyses.
Potential biases in the review process
One of the problems involved with merging the data for this review may be generalisability.
In our review, most participants had a presumed cardiac cause of cardiac arrest while non‐cardiac aetiologies of out‐of‐hospital cardiac arrest may account for 30% to 40% of all cases (Engdahl 2002; Hawkes 2017; Kuisma 1997). This may have resulted in an under‐representation of adults with cardiac arrest of non‐cardiac origin.
Mori 2000 was published as an abstract. We were able to receive additional information on the cooling method from the authors but no details on some quality criteria for this RCT. For most quality criteria we had 'no information', which resulted in an overall assessment of high risk of bias. Bernard 2002 was a quasi‐RCT, which was the main reason for it having a high risk of bias. The overall effect was slightly reduced by excluding those studies from the analysis.
Information on the important aspects of study quality, such as the details of randomisation and information of a prepublished analysis protocol were not available for many studies, especially the older ones. For some studies, we were unable to retrieve more information (e.g. on quality parameters from Mori 2000, or dichotomous outcome data from Zhang 2005).
We included the data from Zhang 2005 by transforming the continuous data to dichotomised data (see Altman 1991), according to the cut‐off point for favourable versus unfavourable neurological outcomes provided by the study authors. There are limitations of the method (loss of information, assumption of normality). We added a sensitivity analysis excluding the data, which resulted in a slight reduction of the effect (from RR 1.41 to 1.33, 95% CI 1.07 to 1.65).
Studies reported information on cardiac arrest in a variety of ways and details. For example, details on the no‐flow time (duration from cardiac arrest to start of resuscitation) is a parameter that is difficult to document reliably and at the same time crucial for the outcome, and only four studies reported it (Bernard 2002; Kwon 2021; Laurent 2005; Nielsen 2013). A more uniform reporting of characteristics of cardiac arrest as well as the intervention (timing of cooling) would have enabled a better assessment of the groups of participants and more powerful subgroup analyses.
We have included several factors of the cardiac arrest situation and the intervention to provide a more granular picture of the effectiveness of hypothermia in certain subgroups. The interval between ROSC and the initiation of cooling is a crucial factor that signifies the timely implementation of the intervention. Five RCTs did not report this variable (Dankiewicz 2021; Hachimi‐Idrissi 2005 SSP; Mori 2000; Nielsen 2013; Zhang 2005). Therefore, we have included a subgroup analysis on all studies with a reported interval from ROSC to cooling (less than two hours) and studies that did not report on this variable.
Whether the TTM studies are comparable to the other studies is an ongoing discussion in the scientific community. Instead of presenting subgroup analyses or an analysis of the overall estimate only, we present both approaches in our review to allow for a transparent presentation of the evidence in light of potential sources of heterogeneity. The trials by the TTM study group were the largest in this field (Dankiewicz 2021; Nielsen 2013). However, they were pragmatic in several aspects (e.g. multicentre pragmatic design, the intervention included various methods of cooling), which adds the efficacy‐effectiveness gap issue to the interpretation of the findings in relation to the other studies. Duration from collapse to resuscitation was very short (one minute) compared to the other trials (approximately 10 minutes). Moreover, they had a higher rate of bystander CPR (82% in Dankiewicz 2021 compared to approximately 50% in Bernard 2002, HACA 2002, and Kwon 2021). Previous research showed that only after three minutes or more of no‐flow time hypothermia may be beneficial (Testori 2012). Therefore, the target population may have differed between the subgroups, too.
Agreements and disagreements with other studies or reviews
Several systematic reviews and meta‐analyses have been published, the ones incorporating the most recent data from larger trials are Aneman 2022, Fernando 2021, Granfeldt 2021, and Elbadawi 2022.
Fernando 2021 presented the results of a network meta‐analysis (NMA) comparing the efficacy and safety of deep hypothermia (body temperature 31 °C to 32 °C), moderate hypothermia (33 °C to 34 °C), mild hypothermia (35 °C to 36 °C), and normothermia (37 °C to 37.8 °C) during TTM. They searched six bibliographic databases with 15 search terms and included 10 trials. The authors concluded that none of the hypothermia groups in comparison to normothermia improved survival or survival with good neurological function. Four trials that would be eligible were not included in the review (Hachimi‐Idrissi 2005 LSP; Hachimi‐Idrissi 2005 SSP; Kwon 2021; Mori 2000; Zhang 2005). As for the study design, an NMA is useful as it may enable the comparison of direct and indirect effects between different target temperatures; however, certain prerequisites have to be met. The validity of an indirect comparison requires that the different sets of RCTs are similar in all important factors other than the intervention comparison being made (Chaimani 2022). The review by Fernando 2021 included studies with different study populations (e.g. study populations with cardiac arrest of shockable versus non‐shockable rhythm, populations with a high rate of bystander CPR versus populations with low rates of bystander CPR, and studies with different control group treatments). Incoherence may have been an issue and may have compromised the validity of the analysis.
The second systematic review and meta‐analysis is important as it was undertaken by ILCOR Advanced Life Support (ALS) Task Force to serve as the basis for the ILCOR Consensus on Science with Treatment Recommendations (CoSTR) on temperature management in adult cardiac arrest (Soar 2021). The aim was to "perform a systematic review and meta‐analysis of all aspects of TTM including timing, temperature, duration, method, and rewarming to inform international cardiac arrest guidelines". The authors searched three bibliographic databases and the International Clinical Trials Registry Platform (ICTRP) with an updated search in June 2021. They retrieved 2328 unique records. The authors concluded that the use of TTM at 32 °C to 34 °C when compared to normothermia, did not result in improved outcomes. The conclusion is different to the conclusion in our review. Possible reasons are:
-
the authors analysed many subgroups of different interventions and time points of the outcome but did not provide a summary meta‐analysis of all available evidence. The trials by Hachimi‐Idrissi 2001, and Hachimi‐Idrissi 2005 SSP were not included, the authors justified this by a shorter cooling period; however, this criterion was not mentioned in the protocol of the review or the methods section. Studies published as abstracts only were excluded, hence Mori 2000 was not considered for the meta‐analysis;
-
the partitioning of the available data into subgroups of different time points of outcomes resulted in several meta‐analyses with few studies and low power. In our review, we did not divide the data between different time points, as studies of survivors of cardiac arrest have shown that hypothermia after cardiac arrest was not associated with a change in outcome beyond one month (Arrich 2009b). Also, mortality measured at different time points did not influence pooled point estimates of the effects in critical care trials (Roth 2016);
-
one study that was published later in 2021 could not be included in the reivew (i.e. Kwon 2021);
The review of Aneman 2022 evaluated the effects of hypothermia for at least 12 hours in comparison to no hypothermia in adult survivors of cardiac arrest. The search strategy was adequate and identified 4201 studies. They included seven RCTs and concluded that the posterior probability distributions did not support the use of TTM at 32 °C to 34 °C compared to 36 °C, also including active control of fever to reduce the risk of death and unfavourable neurological outcome at 90 to 180 days. The analysis did not include the studies of Mori 2000, Kwon 2021, and Zhang 2005 (reasons not stated). According to the eligibility criteria of the review, all studies with shorter cooling durations were excluded (i.e. Hachimi‐Idrissi 2001; Hachimi‐Idrissi 2005 SSP). This may have resulted in a more selected view of the full available evidence on the topic.
The meta‐analysis by Elbadawi 2022 aimed at evaluating the outcomes of targeted hypothermia versus normothermia in comatose people after cardiac arrest. The criterion for the control group was "normothermia" but the temperature targets were not defined. It was unclear if the control group was allowed to receive any temperature management. The analysis did not include the trials by Kwon 2021, Mori 2000, Nielsen 2013, and Zhang 2005, partly due to the eligibility criteria. The authors concluded there were no improved outcomes with TTM.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Forest plot: conventional cooling versus control: 1.1 neurological outcome.
Forest plot: haemofiltration cooling versus haemofiltration normothermia: 1.2: neurological outcome.
Forest plot: conventional cooling versus control: 1.2: survival.
Forest plot: haemofiltration cooling versus haemofiltration normothermia: 2.2 survival.
Comparison 1: Neurological outcome: therapeutic hypothermia versus control, Outcome 1: Neurological outcome: conventional cooling versus control, all studies
Comparison 1: Neurological outcome: therapeutic hypothermia versus control, Outcome 2: Neurological outcome: haemofiltration cooling versus haemofiltration normothermia
Comparison 2: Survival: therapeutic hypothermia versus control, Outcome 1: Survival: conventional cooling versus control, all studies
Comparison 2: Survival: therapeutic hypothermia versus control, Outcome 2: Survival: cooling with haemofiltration versus normothermia with haemofiltration
Comparison 3: Neurological outcome: subgroup analyses of cardiac arrest conditions, Outcome 1: Cause of cardiac arrest
Comparison 3: Neurological outcome: subgroup analyses of cardiac arrest conditions, Outcome 2: Location of cardiac arrest
Comparison 3: Neurological outcome: subgroup analyses of cardiac arrest conditions, Outcome 3: Primary cardiac rhythm
Comparison 3: Neurological outcome: subgroup analyses of cardiac arrest conditions, Outcome 4: Witnesses cardiac arrest
Comparison 3: Neurological outcome: subgroup analyses of cardiac arrest conditions, Outcome 5: Bystander cardiopulmonary resuscitation rate
Comparison 3: Neurological outcome: subgroup analyses of cardiac arrest conditions, Outcome 6: No‐flow time
Comparison 3: Neurological outcome: subgroup analyses of cardiac arrest conditions, Outcome 7: Duration of hypothermia
Comparison 3: Neurological outcome: subgroup analyses of cardiac arrest conditions, Outcome 8: Time interval from return of spontaneous circulation (ROSC) to intervention
| Effects of cooling 32–34 °C on neuroprotection, survival (to discharge or within 6 months) and adverse events for comatose survivors of cardiac arrest | ||||||||||
| Patient or population: adults after cardiopulmonary resuscitation Settings: emergency medicine and intensive care, worldwide Intervention: cooling 32–34 °C (good neurological outcome and survival); cooling via haemofiltration (adverse events) Control: control treatment at ≥ 36 °C | ||||||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect (95% CI) | Number of participants (studies) | Risk of bias | Inconsistency | Indirectness | Imprecision | Other considerations | Overall certainty of evidence | |
|---|---|---|---|---|---|---|---|---|---|---|
| Risk with control | Risk difference | |||||||||
| ≥ 36 °C | Hypothermia at 32–34 °C | |||||||||
| Good neurological outcome | Study population | RR 1.41 | 3914 (11) | Seriousa | Seriousb | Not serious | Not serious | None | ⊕⊕⊖⊖ | |
| 377 per 1000 | 155 more per 1000 | |||||||||
| Survival | Study population | RR 1.07 | 3871 | Seriousa | Seriousb | Not serious | Not serious | None | ⊕⊕⊖⊖ | |
| 448 per 1000 | 31 more per 1000 | |||||||||
| Adverse events – pneumonia | Study population | RR 1.09 | 3634 (4 studies) | Seriousa | Not serious | Not serious | Seriousc | None | ⊕⊕⊖⊖ | |
| 352 per 1000 | 32 more per 1000 | |||||||||
| Adverse events – hypokalaemia | Study population | RR 1.38 | 975 (2 studies) | Seriousa | Not serious | Seriousd | Seriousc | None | ⊕⊖⊖⊖ | |
| 134 per 1000 | 51 more per 1000 | |||||||||
| Adverse events – arrhythmia | Study population | RR 1.40 | 2163 (3 studies) | Seriousa | Not serious | Seriousd | Not serious | None | ⊕⊕⊖⊖ | |
| 184 per 1000 | 73 more per 1000 | |||||||||
| *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and therelative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio. | ||||||||||
| GRADE Working Group grades of evidence | ||||||||||
| aDowngraded one level as there were at least 'some concerns' for all studies (Bernard 2002 and Mori 2000 had high risk of bias). | ||||||||||
| Outcome or subgroup | Studies | Participants | Risk ratio (95% CI) |
|---|---|---|---|
| Bleeding of any severity | 4 | 3636 | 1.09 (0.94 to 1.27) |
| Need for platelet transfusion | 1 | 273 | 5.11 (0.25 to 105.47) |
| Significant haemorrhagic complications | 1 | 77 | Not estimable |
| Pneumonia | 4 | 3634 | 1.09 (1.00 to 1.18) |
| Pancreatitis | 1 | 273 | 0.51 (0.05 to 5.57) |
| Sepsis | 3 | 3054 | 1.17 (0.94 to 1.45) |
| Septic shock | 1 | 933 | 0.87 (0.50 to 1.52) |
| Renal failure or oliguria | 2 | 303 | 0.88 (0.48 to 1.61) |
| Haemodialysis | 4 | 1869 | 1.12 (0.85 to 1.48) |
| Seizures | 3 | 1783 | 1.11 (0.95 to 1.30) |
| Severe, haemodynamic compromising or long‐lasting arrhythmia | 3 | 2163 | 1.40 (1.19 to 1.64) |
| Any arrhythmia | 1 | 933 | 0.98 (0.93 to 1.04) |
| Pulmonary oedema | 2 | 850 | 0.93 (0.57 to 1.52) |
| Cardiac complications | 1 | — | No totals |
| Hypokalaemia | 2 | 975 | 1.38 (1.03 to 1.84) |
| Hypophosphataemia | 2 | 975 | 1.10 (0.92 to 1.33) |
| Hypoglycaemia | 1 | 933 | 1.12 (0.64 to 1.97) |
| Hypomagnesaemia | 1 | 933 | 1.20 (0.88 to 1.65) |
| Pressure sores | 1 | 269 | Not estimable |
| Skin complications related to device | 1 | 1849 | 1.99 (0.68 to 5.80) |
| Bacteraemia | 1 | 581 | 1.14 (0.51 to 2.54) |
| Central venous catheter infection | 1 | 581 | 2.09 (0.39 to 11.33) |
| Urinary tract infections | 1 | 581 | 0.78 (0.34 to 1.83) |
| Nosocomial infections, other than central venous catheter infection and urinary tract infections | 1 | 581 | 1.05 (0.31 to 3.57) |
| Ventilator‐associated pneumonia | 1 | 581 | 1.31 (0.87 to 1.99) |
| Vasopressors between day 0 and 7 | 1 | 581 | 1.01 (0.94 to 1.09) |
| CI: confidence interval. | |||
| Sensitivity analysis | Measure | Risk ratio (95% CI) |
|---|---|---|
| Robustness against risk of bias in individual studies | Excluding studies with high risk of bias (Bernard 2002; Mori 2000) | 1.29 (1.04 to 1.60) |
| Robustness against methods of the intervention | Pooling with studies using haemofiltration (Laurent 2005) | 1.35 (1.09 to 1.67) |
| Robustness against model choice | Using fixed‐effect model instead of random‐effects model | 1.09 (1.02 to 1.18) |
| Robustness against including converted data | Excluding converted study data from a continuous outcome (Zhang 2005) | 1.33 (1.07 to 1.65) |
| CI: confidence interval. | ||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1.1 Neurological outcome: conventional cooling versus control, all studies Show forest plot | 11 | 3914 | Risk Ratio (M‐H, Random, 95% CI) | 1.41 [1.12, 1.76] |
| 1.1.1 Neurological outcome: conventional cooling versus no cooling or fever control | 8 | 2870 | Risk Ratio (M‐H, Random, 95% CI) | 1.60 [1.15, 2.23] |
| 1.1.2 Neurological outcome: conventional cooling versus 36 °C | 3 | 1044 | Risk Ratio (M‐H, Random, 95% CI) | 1.78 [0.70, 4.53] |
| 1.2 Neurological outcome: haemofiltration cooling versus haemofiltration normothermia Show forest plot | 1 | 42 | Risk Ratio (M‐H, Random, 95% CI) | 0.71 [0.32, 1.54] |
| 1.2.1 Neurologic outcome: haemofiltration cooling versus haemofiltration normothermia | 1 | 42 | Risk Ratio (M‐H, Random, 95% CI) | 0.71 [0.32, 1.54] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 2.1 Survival: conventional cooling versus control, all studies Show forest plot | 9 | 3871 | Risk Ratio (M‐H, Random, 95% CI) | 1.07 [0.95, 1.20] |
| 2.1.1 Survival: therapeutic hypothermia with conventional cooling methods versus control | 7 | 2875 | Risk Ratio (M‐H, Random, 95% CI) | 1.17 [0.96, 1.42] |
| 2.1.2 Survival: conventional cooling versus 36 °C | 2 | 996 | Risk Ratio (M‐H, Random, 95% CI) | 0.98 [0.86, 1.10] |
| 2.2 Survival: cooling with haemofiltration versus normothermia with haemofiltration Show forest plot | 1 | 42 | Risk Ratio (M‐H, Random, 95% CI) | 0.71 [0.32, 1.54] |
| 2.2.1 Survival: haemofiltration cooling versus haemofiltration normothermia | 1 | 42 | Risk Ratio (M‐H, Random, 95% CI) | 0.71 [0.32, 1.54] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 3.1 Cause of cardiac arrest Show forest plot | 4 | Risk Ratio (M‐H, Random, 95% CI) | Subtotals only | |
| 3.1.1 Non‐cardiac cause | 3 | 434 | Risk Ratio (M‐H, Random, 95% CI) | 1.01 [0.45, 2.27] |
| 3.1.2 Cardiac cause | 4 | 1386 | Risk Ratio (M‐H, Random, 95% CI) | 1.27 [0.89, 1.82] |
| 3.2 Location of cardiac arrest Show forest plot | 10 | 3897 | Risk Ratio (M‐H, Random, 95% CI) | 1.34 [1.09, 1.65] |
| 3.2.1 Out‐of‐hospital cardiac arrest | 10 | 3721 | Risk Ratio (M‐H, Random, 95% CI) | 1.28 [1.04, 1.57] |
| 3.2.2 In‐hospital cardiac arrest | 2 | 176 | Risk Ratio (M‐H, Random, 95% CI) | 2.29 [1.05, 4.99] |
| 3.3 Primary cardiac rhythm Show forest plot | 7 | 1939 | Risk Ratio (M‐H, Random, 95% CI) | 1.36 [1.08, 1.72] |
| 3.3.1 Ventricular fibrillation/ventricular tachycardia/shockable rhythms | 4 | 1087 | Risk Ratio (M‐H, Random, 95% CI) | 1.38 [1.02, 1.87] |
| 3.3.2 Asystole/pulseless electrical activity/non‐shockable rhythms | 5 | 852 | Risk Ratio (M‐H, Random, 95% CI) | 1.64 [0.84, 3.21] |
| 3.4 Witnesses cardiac arrest Show forest plot | 5 | 991 | Risk Ratio (M‐H, Random, 95% CI) | 1.53 [1.24, 1.90] |
| 3.4.1 Witnessed cardiac arrest | 5 | 935 | Risk Ratio (M‐H, Random, 95% CI) | 1.48 [1.19, 1.83] |
| 3.4.2 Non‐witnessed cardiac arrest | 4 | 56 | Risk Ratio (M‐H, Random, 95% CI) | 5.52 [1.62, 18.86] |
| 3.5 Bystander cardiopulmonary resuscitation rate Show forest plot | 9 | 3844 | Risk Ratio (M‐H, Random, 95% CI) | 1.27 [1.04, 1.54] |
| 3.5.1 ≥ 70% | 3 | 3343 | Risk Ratio (M‐H, Random, 95% CI) | 1.02 [0.89, 1.18] |
| 3.5.2 43–59% | 3 | 378 | Risk Ratio (M‐H, Random, 95% CI) | 1.47 [1.17, 1.86] |
| 3.5.3 7–42% | 3 | 123 | Risk Ratio (M‐H, Random, 95% CI) | 2.91 [1.30, 6.55] |
| 3.6 No‐flow time Show forest plot | 4 | 1844 | Risk Ratio (M‐H, Random, 95% CI) | 1.31 [0.94, 1.84] |
| 3.6.1 Up to 1 min | 1 | 933 | Risk Ratio (M‐H, Random, 95% CI) | 0.97 [0.85, 1.11] |
| 3.6.2 1–2 min | 1 | 581 | Risk Ratio (M‐H, Random, 95% CI) | 1.78 [1.00, 3.17] |
| 3.6.3 3–5 min | 2 | 330 | Risk Ratio (M‐H, Random, 95% CI) | 1.43 [1.12, 1.84] |
| 3.7 Duration of hypothermia Show forest plot | 11 | 3914 | Risk Ratio (M‐H, Random, 95% CI) | 1.41 [1.12, 1.76] |
| 3.7.1 4 hours | 2 | 66 | Risk Ratio (M‐H, Random, 95% CI) | 4.42 [1.26, 15.58] |
| 3.7.2 12–28 hours | 7 | 3778 | Risk Ratio (M‐H, Random, 95% CI) | 1.20 [1.00, 1.45] |
| 3.7.3 72 hours | 2 | 70 | Risk Ratio (M‐H, Random, 95% CI) | 3.79 [1.64, 8.73] |
| 3.8 Time interval from return of spontaneous circulation (ROSC) to intervention Show forest plot | 11 | 3914 | Risk Ratio (M‐H, Random, 95% CI) | 1.41 [1.12, 1.76] |
| 3.8.1 Intervention started within 2 hours of ROSC | 6 | 1049 | Risk Ratio (M‐H, Random, 95% CI) | 1.57 [1.27, 1.94] |
| 3.8.2 Time interval from ROSC to intervention not reported | 5 | 2865 | Risk Ratio (M‐H, Random, 95% CI) | 1.10 [0.87, 1.38] |
| Bias | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Randomisation process | Deviations from intended interventions | Missing outcome data | Measurement of the outcome | Selection of the reported results | Overall | ||||||
| Subgroup 1.1.1 Neurological outcome: conventional cooling versus no cooling or fever control | ||||||||||||
| Bernard 2002 |  |  |  | | | | ||||||
| Quasi randomized trial | ||||||||||||
| 5 out of 43 patients in the hypothermia group did not receive the intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Quasi randomisation, result of overall RoB 2 assessment | ||||||||||||
| Dankiewicz 2021 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention | ||||||||||||
| HACA 2002 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention | ||||||||||||
| Hachimi‐Idrissi 2001 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Study protocol was not registered. However, outcomes were prespecified in study protocol and in accordance with the reporting recommendations | ||||||||||||
| Result of RoB 2 algorithm | ||||||||||||
| Hachimi‐Idrissi 2005 LSP | | | | | | | ||||||
| The method of randomisation was not described in this study but was described in other RCTs of the study group. | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Hachimi‐Idrissi 2005 SSP | | | | | | | ||||||
| The method of randomisation was not described in this study but was described in other RCTs of the study group. | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Lascarrou 2019 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Zhang 2005 | | | | | | | ||||||
| Method of randomisation not described | ||||||||||||
| Carers and people delivering the interventions were aware of participants' assigned intervention during the trial | ||||||||||||
| No missing outcome data | ||||||||||||
| No information on blinding of outcome assessors | ||||||||||||
| No information on a prespecified analysis plan. | ||||||||||||
| No information on details of randomisation, blinded outcome assessment, and analysis plan. | ||||||||||||
| Subgroup 1.1.2 Neurological outcome: conventional cooling versus 36 °C | ||||||||||||
| Kwon 2021 | | | | | | | ||||||
| More patients in the TTM 33 °C group received bystander CPR, but numbers were small and difference could have been due to chance. | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Mori 2000 | | | | | | | ||||||
| No information on method of randomisation or allocation concealment available. | ||||||||||||
| No information on blinding of carers and people delivering the intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| No information on the protocol was available. However, outcomes were in agreement with recommended reporting guidelines in cardiac arrest studies. | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Nielsen 2013 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Bias | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Randomisation process | Deviations from intended interventions | Missing outcome data | Measurement of the outcome | Selection of the reported results | Overall | ||||||
| Subgroup 1.2.1 Neurologic outcome: haemofiltration cooling versus haemofiltration normothermia | ||||||||||||
| Laurent 2005 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| No information if outcome assessors were aware of the intervention received by study participants. | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Bias | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Randomisation process | Deviations from intended interventions | Missing outcome data | Measurement of the outcome | Selection of the reported results | Overall | ||||||
| Subgroup 2.1.1 Survival: therapeutic hypothermia with conventional cooling methods versus control | ||||||||||||
| Bernard 2002 | | | | | | | ||||||
| Quasi randomized trial | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention | ||||||||||||
| 5 out of 43 patients in the hypothermia group did not receive the intervention. | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Quasi randomization | ||||||||||||
| Dankiewicz 2021 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| HACA 2002 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Hachimi‐Idrissi 2001 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Hachimi‐Idrissi 2005 LSP | | | | | | | ||||||
| The method of randomisation was not described in this study but was described in other RCTs of the study group. | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Hachimi‐Idrissi 2005 SSP | | | | | | | ||||||
| The method of randomisation was not described in this study but was described in other RCTs of the study group. | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Lascarrou 2019 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Subgroup 2.1.2 Survival: conventional cooling versus 36 °C | ||||||||||||
| Kwon 2021 | | | | | | | ||||||
| More patients in the TTM 33 °C group received bystander CPR, but numbers were small and difference could have been due to chance. | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Nielsen 2013 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
| Bias | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Randomisation process | Deviations from intended interventions | Missing outcome data | Measurement of the outcome | Selection of the reported results | Overall | ||||||
| Subgroup 2.2.1 Survival: haemofiltration cooling versus haemofiltration normothermia | ||||||||||||
| Laurent 2005 | | | | | | | ||||||
| Adequate | ||||||||||||
| Carers and people delivering the intervention were aware of the participants' assigned intervention. | ||||||||||||
| Adequate | ||||||||||||
| No information if outcome assessors were aware of the intervention received by study participants. | ||||||||||||
| Adequate | ||||||||||||
| Result of overall RoB 2 assessment | ||||||||||||
