Therapeutic Hypothermia Following Cardiopulmonary Arrest: A Systematic Review and Meta-Analysis with Trial Sequential Analysis

The primary outcome of neurologic outcome was available for 3611 subjects. Of the included trials, four trials used the Cerebral Performance Category (CPC) or similar scale for determining neurologic outcomes [2,3,5,6]. The remaining trial utilized the modified Rankin Scale (mRS)[4] A follow up duration of six months was used for the favorable outcome in three studies[2,4,5]. The remaining two studies used hospital discharge [3] and 90 days [6] for their outcome follow up period. Of the included studies, 38.7% (702/1813) and 36.5% (657/1798) of the participants had a favorable outcome in the hypothermia and normothermia groups, respectively (Table S1). The random effects meta-analysis revealed no significant difference between the two groups in favorable neurologic outcome (risk ratio [RR] 0.94, 95% confidence interval [CI] 0.87 to 1.03) with a heterogeneity of 63% (Figure 1).

Fig. 1.

A total of 3613 subjects from the five trials had follow up data available for all-cause mortality. Longest follow up for all-cause mortality mirrored that of the longest follow up for neurologic outcomes, with majority of studies reporting the outcome at six months. Overall, mortality was 54.2% (1009/1862) and 54.7% (1017/1860) in the hypothermia and normothermia groups, respectively (Table S1). There was no significant difference in all-cause mortality between groups (RR 0.97, 95% CI 0.88 to 1.06) with a heterogeneity of 54% (Figure 2).

Fig. 2.

Definitions for each adverse effect were extracted from each study (Table S2). All except one trial [5] presented data in a per-participant methodology, leaving approximately 2780 subjects available for the analysis of arrhythmias, bleeding, and infection. Seizures were reported at a per-participant level for 927 subjects. Our analysis revealed the hypothermia group had a statistically significant increase in the risk of adverse effects (23.1% vs. 20.3%, RR 1.16, 95% CI 1.04 to 1.28) with a heterogeneity of 47.4% (Figure 3). When adverse effects were assessed by sub-type, arrhythmias occurred in significantly more patients in the hypothermia group compared to the normothermia group (22% vs. 16.3%; RR 1.35; 95% CI 1.16 to 1.57). Other adverse effects including bleeding, seizures, and infections were not different between groups (Figure 3 & Table 2).

Fig. 3.

When comparing the available data of the five trials, three specified subgroups a priori had meaningful data to perform subgroup analyses. Subgroups were divided into the following: initial shockable rhythm (studies with > 80% vs. < 80% initial shockable rhythm), bystander CPR (studies with > 70% vs. < 70% bystander CPR), and duration of cooling (studies with > 24 hours vs. ≤ 24 hours of cooling). Of the subgroup analyses performed, studies with a higher percentage of patients with initial shockable rhythm (Figure 4) were more likely to have a favorable neurologic outcome (53.6% vs. 36.8%, RR 0.73, 95% CI 0.6 to 0.88). The remaining subgroup analyses on neurologic outcome revealed no significant differences between groups and are reported in the supplementary appendices (Figures S4 & S5). Additionally, with the signal of favorable neurologic outcome in trials with larger percentages of patients with shockable rhythms, we conducted a subgroup analysis for the outcome of mortality. Four trials, representing 1685 subjects, reported individual patient data for this analysis [2,3,5,6]. Mortality was not different in this subgroup analysis (RR 0.92, 95% CI 0.8 to 1.05) (Figure S6).

TSA revealed the z-curve for outcomes of favorable neurologic outcome and mortality crossed the lines of futility, signifying randomized controlled trials with these populations would not demonstrate a benefit and would be considered futile (Figures S7-S8). Furthermore, the z-curve for the outcome of adverse effects crossed the line of harm, suggesting that if the systematic review was a randomized trial that it may have been stopped early by an independent data safely monitoring committee due to risk of harm (Figure S9).

Whether therapeutic hypothermia improves survival is controversial, especially if neurologic outcome is not favorable. Our results are similar to previous meta-analyses that demonstrated no survival benefit with hypothermia [18–23,27]. In parallel with the analysis by Sanfilippo and colleagues, our TSA confirmed the futility of further randomized controlled trials in respect to the mortality outcome [22]. All of these findings combined suggest therapeutic hypothermia in the general cardiac arrest population is likely not to provide a survival benefit. However, the question regarding benefit in select populations remains, and our subgroup analysis of individual patient level data showed a non-significant trend towards improved survival. Even though neurologic outcome showed benefit with hypothermia in the shockable rhythm subgroup analysis, mortality was not different leading to discrepancies of which outcome is most important.

This outcome was also associated with limitations necessary to note. Each of these trials reported mortality at variable follow up durations, which ranged from hospital discharge to six months. Although unlikely to confound our findings with the use of a random effects model, it is important to note these outcomes were measured at differing time points. Additionally, another important limitation was for our subgroup analysis on shockable rhythms. The recent randomized control trial by Dankiewicz and colleagues did not report individual patient data in regards to mortality and shockable rhythms [4]. Although, this would have been unlikely to influence our findings as the reported confidence interval for this subgroup in the trial demonstrated no difference.

Adverse effects are important when considering any therapeutic strategy to determine the risk to benefit ratio for the patient. Therapeutic hypothermia is not a benign therapy and has been associated with various adverse effects. Few of the previous meta-analyses have reported on adverse effect findings; however, of those referenced, patients randomized to therapeutic hypothermia were at an increased risk for arrhythmias [18,19,21,22]. Our study revealed consistent results in comparison to previous meta-analyses, with therapeutic hypothermia patients having a 35% increased relative risk of experiencing an arrhythmia. In contrast to prior meta-analyses, our study provided a thorough picture of the potential adverse effects of therapeutic hypothermia through examining a comprehensive list of adverse effects and their respective incidences, including bleeding (6.8%), infections (42.7%), and seizures (16.6%). Although we found no independent differences in these outcomes compared to the normothermia group, when compiled together, patients in the therapeutic hypothermia group remained at a 16% increased relative risk for adverse effects. To our knowledge, our study is the first to perform a TSA on adverse effect outcomes, which revealed futility of therapeutic hypothermia.

Important limitations exists in regards to adverse effect reporting and our analysis. First, each trial utilized their own method for reporting, with some trials using an independent data safety monitoring committee, while others were subjected to physician reporting. Due to each trial using their own definitions for adverse effects, we were unable to delineate the predominant arrhythmia reported. Similarly, the standard of care for adverse effect reporting has changed over time and become more objective according to the requirements of independent data safety monitoring committees; however, our analysis with TSA and alpha spending is aimed to adjust for this change, as mentioned previously. Additionally, one trial reported adverse effects over a total of reported events, rather than per participant, hindering our ability to include those adverse effects in our initial subgroup analysis [5]. Administration of prophylactic lidocaine therapy was performed in one of the trials and could have led to an under-reporting of arrhythmias [3]. Additionally, pre-existing structural heart disease and post-cardiac arrest myocardial dysfunction predispose patients to arrhythmias and may have contributed to the reported adverse effects in both groups [28]. Other adverse effects of hypothermia are also important to note and were not consistently reported in these studies, which include metabolic disorders and electrolyte abnormalities.

In regards to future directions of randomized controlled trials, TSA revealed futility of further trials in the comparison of hypothermia to normothermia with the outcomes of mortality, neurologic outcomes, and adverse effects. A potential direction for further analyses is prevention of pyrexia in post-cardiac arrest management and re-focusing TTM efforts on patients with shockable rhythms. Pyrexia increases metabolic demand and can lead to consequential cerebral inflammation, increased cerebral blood flow and subsequent elevated intracranial pressure, and neurotransmitter excitotoxicity [29,30]. Following cardiac arrest, occurrence of pyrexia has been associated with unfavorable outcomes [31–34]. Earlier studies with significant survival and neurologic benefit of hypothermia did not actively prevent pyrexia in the control group [2,3]. With no apparent survival or neurologic benefits of therapeutic hypothermia and the risk of adverse effects, perhaps future trial considerations should consist of prevention of pyrexia from a pharmacologic and physiologic perspective in patients with targeted normothermia of less than 38 degrees Celsius.