Hunter BR1,O'donnell DP,Allgood KL,Seupaul RA

1

No benefit to prehospital initiation of therapeutic hypothermia in out-of-hospital cardiac arrest: a systematic review and meta-analysis.

Hunter BR1,O'Donnell DP,Allgood KL,Seupaul RA.

Acad Emerg Med.2014 Apr;21(4):355-64

ABSTRACT:

Objective: The aim of this review is to define the effect of prehospital therapeutic hypothermia (TH) on survival and neurologic recovery in patients who have suffered an out-of-hospital cardiac arrest (OHCA).

Methods: Included in this review are randomized trials assessing the impact of prehospital TH in adult patients suffering a non-traumatic OHCA. Trials assessing the impact of in-hospital TH were excluded. Only studies with a low risk of bias were eligible for meta-analysis. A medical librarian searched PubMed, Ovid, EMBASE, Ovid Global Health, the Cochrane Library, Guidelines.gov, EM Association Websites, CenterWatch, IFPMA Clinical Trial Results Portal, CINAHL, ProQuest, and the Emergency Medical Abstracts Database without language restrictions. Clinicaltrials.gov was searched for unpublished studies. Bibliographies were hand searched and experts in the field were queried about other published or unpublished trials. Using standardized forms, 2 authors independently extracted data from all included trials. Results from high quality trials were pooled using a random effects model. Two authors, using the Cochrane Risk of Bias Tool, assessed risk of Bias independently.

Results: Of 740 citations, 6 trials met inclusion criteria. Four trials were at a low risk of bias and were included in the meta-analysis (N=715). Pooled analysis of these trials revealed no difference in overall survival (RR=0.98, 95% CI= 0.79 to 1.21) or good neurologic outcome (RR=0.96, 95% CI= 0.76 to 1.22) between patients randomized to prehospital TH versus standard therapy. Heterogeneity was low for both survival and neurologic outcome (I2 = 0).

Conclusions: Randomized trial data demonstrates no patient important benefit from prehospital initiation of TH. Pending the results of ongoing larger trials, resources dedicated to this intervention may be better spent elsewhere.

INTRODUCTION

Out of Hospital Cardiac Arrest (OHCA) is a major public health issue. The American Heart Association estimates that approximately 360,000 episodes of OHCA occur each year in the United States,1 with 60%treated by Emergency Medical Services (EMS).1, 2Similarly, the estimated incidence of EMS treated OHCA in Europe is 275,000 annually.3 The overall survival of EMS treated OHCA in adults is estimated to be 9.5%.1 Of those who survive to hospital discharge, the proportion of patients who survive with good neurologic function(Cerebral Performance Category ≤ 2) ranges from 70-90%.4 With such dismal outcomes, resuscitation science continues to focus on finding interventions to improve survival and neurologic function after OHCA.

Therapeutic Hypothermia (TH) has been shown, in randomized trials, to increase both survival and favorable neurologic outcomes in patients suffering from OHCA.5-7 These trials have led the International Liaison Committee on Resuscitation to recommend adoption of TH for patients resuscitated after OHCA.8 Whilein-hospital TH provides proven benefit, it is unclear if initiating this intervention earlier would further improve outcomes. Animal studies suggest that the benefits of TH diminish over time if cooling is delayed.9-12 This has led to interest in achieving TH earlier after OHCA, specifically in the prehospital setting.

Several small non-randomized trials have demonstrated that initiating TH in the prehospital setting is feasible and safe.13-18However, randomized trials have been underpowered to detect differences in clinically relevant outcomes.19, 20Despite this uncertainty, many EMS systems have implemented TH strategies.21 The goal of this systematic review and meta-analysis is to define the effect of prehospital TH on both survival and neurologic recovery in patients who have suffered an OHCA.

METHODS

This systematic review and meta-analysis conforms to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.

Search Strategy

A medical librarian (KA) searched the medical literature from 1946 through December 2012 using the search terms outlined in Appendix A. PubMed, Ovid, EMBASE, Ovid Global Health, the Cochrane Library, Guidelines.gov, EM Association Websites, CenterWatch, IFPMA Clinical Trial Results Portal, CINAHL, ProQuest, and the Emergency Medical Abstracts Database were searched without language restriction. Clinicaltrials.gov was searched for unpublished studies. Experts in the field were queried and bibliographies of relevant trials and reviews were hand searched to identify additional published or unpublished trial data. All titles and abstracts identified by the search were independently screened by 2 authors (RS, BH) for relevance. The full text of all potentially relevant trials was reviewed for inclusion. Disagreements were settled by consensus or adjudication by a third author (DO).

Inclusion/Exclusion Criteria

All randomized, cluster randomized, or quasi-randomized trials comparing prehospital TH to standard prehospital care without TH in adult humanswere included in this systematic review. “Prehospital TH” was defined as any active intervention aimed at lowering the body temperature of patients prior to hospital arrival. Any method of cooling was accepted. OHCA was defined as any non-perfusing cardiac rhythm occurring in a patient not already in or admitted to a hospital. To be included, trials had to report survival to hospital discharge, 28 days, or any period longer than 28 days after cardiac arrest. Non-randomized trials, studies of traumatic cardiac arrest, pediatric arrest, and in hospital arrest were excluded. Studies that randomized patients to in-hospital TH versus no in-hospital TH were also excluded. While all identified trials meeting the inclusion and exclusion criteria are included in the review, only trials with a low risk of bias were eligible to be included in the statistical meta-analysis.

Individual Study Quality Appraisal

Two authors (RS, BH) independently assessed the risk of bias of included trials using standard criteria defined in the Cochrane Handbook for Systematic Reviews of Interventions.22 This validated instrument for appraising randomized trials measures risk of bias in 7 categories: 1) adequate random sequence generation, 2) allocation concealment, 3) blinding of participants and personnel, 4) blinding of outcome assessment, 5) incomplete outcome data, 6) selective reporting, and 7) other bias. Each trial is described as having either a high, low, or unclear risk of bias in each of the 7 domains. Discrepancies were resolved by discussion or adjudication by a third author (DO).

Data Abstraction

Two authors (RS, BH) independently abstracted data using standardized forms. Data abstraction included study design, method of cooling, inclusion and exclusion criteria, number of patients randomized to each arm of the study, number of patients that survived in each arm of the study, and any patients randomized who were not accounted for in the results. Neurologic outcome data, how it was assessed, and number of patients in each group with a “good neurologic outcome” were also abstracted. Additionally, we recorded information regarding known predictors of outcome in cardiac arrest patients: presenting rhythm [ventricular fibrillation (VF) vs. pulseless electrical activity(PEA)/asystole], witnessed arrest, bystander CPR, and whether or not patients were cooled in-hospital.

Outcomes

The primary outcome was survival to hospital discharge or any point at least 28 days post-arrest. If both survival at discharge and at 28+ days was reported, then whichever endpoint came first was used. The secondary outcome was “good neurologic outcome,” defined as a score ≤ 2 for any of the following metrics: Cerebral Performance Category, Overall Performance Category, or the Modified Rankin Score. If one of these validated metrics was not reported, reasonably defined “good neurologic outcome” by the individual study authors was accepted (e.g. Discharged to home/acute rehabilitation, versus discharged to long term care facility/death). We defined,a priori, 2 subgroups for analysis: patients with a presenting rhythm of VF and those with PEA/asystole.

Statistical Analysis

Only trials with a low risk of bias were eligible for meta-analysis (StatsDirect Version 2.7.9). Heterogeneity was assessed using the I2 statistic23 and the X2 test. Data were pooled if statistical heterogeneity was low (I2 50% or the X2p value was > 0.10). Dichotomous variables were reported as relative risks “RR” with 95% confidence intervals “CI”. Data were pooled using a random effects model with sensitivity analyses performed if significant clinical heterogeneity was suspected.

RESULTS:

Search Results

The search identified 740 unique publications. After screening titles and abstracts, 29 full text articles were reviewed, of which 624-29 met inclusion criteria (See Figure 1). Hand searching reference lists and contacting experts in the field did not reveal additional data. No unpublished trials fitting our inclusion criteria were identified.

Risk of Bias of Included Studies

Risk of bias for the 6 included trials is summarized in Table 1. Study quality was variable, and is described below.

Regarding sequence generation and allocation concealment, 4 studies24, 25, 27, 29 were low risk of bias, while 2 trials26, 28gave no description and were scored as havingan unclear risk of bias. Because it was not possible to blind medical personnel to treatment assignment (and no attempts were made to do so), this domain was scored as a high risk of bias for all trials.

Blinding for the assessment of survival was deemed unnecessary; hence all of the included studies were graded as low risk of bias for this outcome. With respect to neurologic outcome, Callaway et al26 was similarly scored as low risk, since no patients survived, providing an objectively poor neurologic outcome for all patients. Both studies by Bernard et al24, 25 were considered to be at low risk of bias for blinding of outcome assessment, while Castren et al27 was graded as high risk based on the admission by the authors that the “assessment may not always have been performed by an individual blinded to the treatment group.”27 The remaining trials were unclear risk28 or did not report neurologic outcomes.29

Risk of bias for incomplete outcome data was low for 4 studies24, 25, 27, 29 and high for 2.26, 28All studies except Kim et al,29which did not report neurologic outcomes, were at low risk of bias for selective outcome reporting.

We assessed all studies for “other risk of bias”. Both studies by Bernard et al24, 25were graded as high risk because they were stopped early due to futility of their 2010 study.24 Castren’s study27 was at unclear risk of bias in that there was participation in the study design, data analysis, and writing of the manuscript by the corporate sponsor. Callaway et al26had a high risk for other bias because their protocol called for cessation of cooling if ROSC was achieved, body temperature of intervention patients was not effectively lowered, and the intervention and control groups appeared prognostically dissimilar at the start of the trial. The remaining trials28, 29 were at low risk for other bias.

Results of Individual Studies and Meta-Analysis:

Methods and results of the included trials are summarized in Table 2.Inclusion and exclusion criteria were generally consistent across studies, with the exception of initial cardiac rhythm (VF vs. non-VF). The most common method of cooling was cold saline infusion.24, 25, 28, 29 Four of the six trials began cooling after ROSC,24, 25, 28, 29 while 2 trials initiated TH intra-arrest.26, 27 None of the 6 trials found a statistically significant difference in survival or neurologic outcomes between patients randomized to prehospital TH versus controls. There were no significant differences between the intervention and control groups in terms of presenting rhythm, bystander CPR, or witnessed arrest, though these were not reported consistently across studies. After quality assessment, 4 trials24, 25, 27, 29were deemed to have a low risk of bias and were included in the meta-analysis. The 2 excluded studies26, 28 had poorly defined methods of randomization and allocation concealment, as well as an excessive number of patients lost to follow up.

Each of the 4 trials included in the pooled analysis24, 25, 27, 29 demonstrated a statistically significant difference in average temperature between intervention patients and controls (-0.8 to -1.3˚C). The average temperature of intervention patients in the 4 trials was 34.2°C to 34.7°C. Pooled analysis of the four trials revealed no significant difference in overall survival (RR=0.98, 95% CI= 0.79 to 1.21) or good neurologic outcome (RR=0.96, 95% CI= 0.76 to 1.22) between patients randomized to prehospital TH versus standard therapy (Figure 2). Heterogeneity was low for both survival and good neurologic outcome (I2 = 0).

Pre-specified subgroup analyses also revealed no differences between groups (Figure 3). Survival data was available from three trials24, 27, 29(n= 344) for patients suffering a VF arrest (RR=1.05, 95% CI=0.77 to 1.44, I2 = 30.8%). Two trials24, 27 (n= 293) reported neurologic outcomesfor patients with VF arrest (RR=1.00 95% CI=0.65 to 1.52).Three trials25, 27, 29 (n= 372) addressedsurvival in patients suffering a non-VF arrest (RR=0.95, 95% CI=0.35 to 2.57, I2 = 46.8%). Neurologic outcomes were reported for non-VF patients in 2 trials25, 27 (n=298; RR=1.30, 95% CI=0.59 to 2.88).

Limitations:

This systematic review and meta-analysis may have several important limitations. None of the included studies blinded healthcare workers to treatment assignment. While this may not have been possible, it could have introduced bias in the form of discrepant co-interventions. It is unlikely, however, that such treatment inequalities would have altered the pooled results. Aside from the lack of blinding, the 4 studies included in our primary analysis24, 25, 27, 29 were at a low risk of bias.

While results across trials were consistent (I2=0), there were important methodological differences between studies (Table 2). The primary differences were initial cardiac rhythm and method and timing of cooling. While 2 trials27, 29 included patients with any rhythm after OHCA, a single study included only patients with VF arrest24 and another included only patients with non-VF arrest.25 Because patients suffering a VF arrest have higher survival rates,27, 29, 30the Bernard et al trial24 disproportionately influenced the pooled results. After performing a sensitivity analysis excluding this trial, however, there remained no statistically significant differences in outcomes.

Of the 4 studies included in the primary analysis, 3 used chilled saline infusion24, 25, 29 to cool patients after ROSC. Castren et al,27 however, relied on coolant infused via an intranasal catheter to cool patients. Castren et al27 was also the only one of the 4 trials to begin the cooling process intra-arrest, as opposed to after ROSC. Despite these differences, Castren’s findings wereconsistent with the other included trials. It is unclear what effect, if any, these differences would have on patient outcomes.

Perhaps even more important than between-study differences in methods of prehospital cooling are the differences between studies in terms of cooling after hospital arrival (Table 2). Prehospital TH would not be expected to provide any clinically important benefit if cooling is not appropriately continued in-hospital. In both studies by Bernard et al,24, 31 all patients in both groups were cooled in-hospital for 24 hours, and Castren’s trial27 states that all patients in both groups were “cooled in the hospital according to institutional standards.” However, Kim et al,29 reported that some receiving hospitals had policies calling for in-hospital TH, while others did not. There was no universal protocol for in-hospital cooling applied to study participants and the authors were not able to provide information on in-hospital utilization of TH. A sensitivity analysis excluding Kim’s trial did not significantly change the results of the meta-analysis. Callaway et al26 called for rewarming after ROSC had been achieved, and Kamarainen et al28 left in-hospital cooling up to the treating physicians. These latter 2 studies26, 28 were not included in the meta-analysis.

Lastly, all of the included studies were performed at large urban centers with short transport times and may not be generalizable to prehospital settings with longer transport times.

Discussion:

There is a growing body of literature examining the impact of prehospital TH. We identified 6 randomized trials,24-29 including 4 high quality studies24, 25, 27, 29 (N=715), assessing the impact of prehospital TH on survival in patients suffering from OHCA. Consistent with the pooled analysis, none of the identified trials demonstrated a statistically significant survival or neurologic outcome benefit.

The prehospital treatment of OHCA continues to evolve. While animal studies9-12 suggest that the benefit of TH diminishes over time with delays in cooling, human observational studies have not demonstrated any consistent effect of delays in achieving TH.32-37 This uncertainty has led to an interest in studying early initiation of TH in the prehospital setting, in the hopes that it might provide added benefit. In the last 10 years, several non-randomized studies have demonstrated the feasibility and safety of TH in the prehospital setting.13-18 The first randomized trial, performed by Callaway et al,26 demonstrated no difference in temperature or survival when applying ice bags around the heads of OHCA victims. In 2007, Kim et al were able to effectively decrease the body temperature of patients after OHCA for the first time in a prehospital randomized trial.29 Although this trial did not find a statistically significant decrease in mortality, there was a trend towards improved survival for patients who had suffered a VF arrest. Despite showing statistical differences in temperature at hospital arrival between intervention and control patients, subsequent randomized trials24, 25, 27, 28 have been unable to demonstrate any statistically significant benefit in survival or neurologic outcomes.

It is unclear why prehospital TH has not been shown to augment the strong evidence supporting in-hospital TH for OHCA patients after ROSC.6, 7 It is possible that current randomized trials have failed to find benefit due inability to achieve meaningful decreases in core temperature in the short time period prior to hospital arrival. Variability in application of in-hospital TH after prehospital initiation is another possible explanation. Absent these potential shortcomings in existing trials, it is also possible that prehospital TH has not proven beneficial because short decreases in the time to initiation or achievement of TH does not provide clinically meaningful benefit.

If TH is not continued in-hospital after prehospital initiation, no benefit would be expected. Three of the 4 trials24, 27, 31 included in this meta-analysis followed protocols for in-hospital TH for all enrolled patients, and exclusion of the trial that did not29 had no appreciable effect on the pooled results. This makes variability between in-hospital therapies an unlikely explanation for the lack of benefit demonstrated in the meta-analysis.

In the studies included in our meta-analysis,24, 27, 29, 31 the difference in average temperature upon ED arrival between patients cooled in the field and controls was modest, ranging from -0.8˚C to -1.3˚C. None of these trials achieved an average temperature of 32-34˚C in intervention patients prior to ED arrival. More rapid and effective means of cooling in the field, and the subsequent impact on patient outcomes, is an area for further research. What this review and meta-analysis does demonstrate, however, is that using currently available methods for inducing TH in the prehospital setting modestly decreases average temperature at ED arrival, without any evidence of improvement in patient important outcomes.