Associations of therapeutic hypothermia with clinical outcomes in patients receiving ECPR after cardiac arrest: systematic review with meta-analysis

Background Therapeutic hypothermia has been recommended for eligible patients after cardiac arrest (CA) in order to improve outcomes. Up to now, several comparative observational studies have evaluated the combined use of extracorporeal cardiopulmonary resuscitation (ECPR) and therapeutic hypothermia in adult patients with CA. However, the effects of therapeutic hypothermia in adult CA patients receiving ECPR are inconsistent. Methods Relevant studies in English databases (PubMed, ISI web of science, OVID, and Embase) were systematically searched up to September 2019. Odds ratios (ORs) from eligible studies were extracted and pooled to summarize the associations of therapeutic hypothermia with favorable neurological outcomes and survival in adult CA patients receiving ECPR. Results 13 articles were included in the present meta-analysis study. There were nine studies with a total of 806 cases reporting the association of therapeutic hypothermia with neurological outcomes in CA patients receiving ECPR. Pooling analysis suggested that therapeutic hypothermia was significantly associated with favorable neurological outcomes in overall (N = 9, OR = 3.507, 95%CI = 2.194–5.607, P < 0.001, fixed-effects model) and in all subgroups according to control type, regions, sample size, CA location, ORs obtained methods, follow-up period, and modified Newcastle Ottawa Scale (mNOS) scores. There were nine studies with a total of 806 cases assessing the association of therapeutic hypothermia with survival in CA patients receiving ECPR. After pooling the ORs, therapeutic hypothermia was found to be significantly associated with survival in overall (N = 9, OR = 2.540, 95%CI = 1.245–5.180, P = 0.010, random-effects model) and in some subgroups. Publication bias was found when evaluating the association of therapeutic hypothermia with neurological outcomes in CA patients receiving ECPR. Additional trim-and-fill analysis estimated four “missing” studies, which adjusted the effect size to 2.800 (95%CI = 1.842–4.526, P < 0.001, fixed-effects model) for neurological outcomes. Conclusions Therapeutic hypothermia may be associated with favorable neurological outcomes and survival in adult CA patients undergoing ECPR. However, the result should be treated carefully because it is a synthesis of low-level evidence and other limitations exist in present study. It is necessary to perform randomized controlled trials to validate our result before considering the result in clinical practices.


Background
Cardiac arrest (CA) remains the leading cause of sudden death and is associated with high mortality despite much progress in advanced life support [1][2][3][4][5][6]. Cardiopulmonary resuscitation using extracorporeal membrane oxygenation (ECPR), also called extracorporeal life support (ECLS), is a modified form of cardiopulmonary bypass. As an alternative resuscitation method, it facilitates the establishment of normal circulation, provides adequate organ perfusion, and then improves both neurological outcomes and survival following CA [7]. Compared with conventional cardiopulmonary resuscitation (CCPR), ECPR is associated with 13% absolute increase of 30 days survival rate in both out-of-hospital cardiac arrest (OHCA) and in-hospital cardiac arrest (IHCA) [7][8][9][10][11][12]. Technological improvements have made it more accessible and its use has been increased over the past decades, especially in patients with refractory CA [13,14]. According to the American Heart Association 2015 Guidelines, ECPR may be considered for selected CA patients with potentially reversible etiology [15].
Despite ECPR, the survival rate at hospital discharge with favorable neurological outcomes remains low in CA patients [16][17][18]. Targeted temperature management (TTM) or mild therapeutic hypothermia via surface cooling has been showed to improve survival and neurological outcomes in patients resuscitated from CA [19,20]. Cooling the body to 32-34°C leads to a relative 35% increase in survival compared with no intervention [21]. International guidelines strongly recommend initiating therapeutic hypothermia for eligible CA patients to improve outcomes [22][23][24]. ECPR is an ideal tool for rapid and homogenous cooling and can also be augmented with surface cooling. Up to now, there are some studies evaluating the combined use of ECPR and therapeutic hypothermia in adult CA patients and/or comparing therapeutic hypothermia treatment with no therapeutic hypothermia induction [9,[25][26][27][28][29][30][31][32][33][34][35][36][37]. However, the sample size of those studies is limited, the proportion of patients receiving therapeutic hypothermia ranges variously, and no conclusive result is derived whether there is a benefit of therapeutic hypothermia treatment in CA patients undergoing ECPR.
In the present study, for the first time, we performed a systematic review with meta-analysis to compare neurological outcomes and survival (Outcomes) between therapeutic hypothermia treatment at 32-34°C (Intervention) and any other temperature controls including no therapeutic hypothermia induction and alternative targeted temperatures range (> 34°C, ≤36°C) (Control) in adult CA patients receiving ECPR (Population).

Materials and methods
The results of this meta-analysis were written according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) reporting guidelines for randomized trials and MOOSE (Meta-analysis Of Observational Studies in Epidemiology) reporting guidelines for observational studies. We used the PICO (Participants/ Population, Intervention/Exposure, Comparison/Control, and Outcomes) scheme to generate the research questions, create a search strategy, and guide study selection. Participants/Population, adult CA patients undergoing ECPR; no restrictions were applied to downtime, bystander CPR attempt, or witnessed arrest. Intervention/ Exposure, therapeutic hypothermia with a targeted temperature of 32-34°C. Comparison/Control, any temperature controls (> 34°C) including no therapeutic hypothermia induction and alternative targeted temperatures range (> 34°C, ≤36°C). Outcomes, [1] favorable neurological outcome at hospital discharge/30 days/90 days, which were evaluated by the cerebral performance category (CPC) with CPC1-2 as favorable outcomes [2]; survival at hospital discharge/30 days/90 days.

Publication search
A systematic search of relevant articles on the associations of therapeutic hypothermia with clinical outcomes (survival or neurological outcomes) in CA patients undergoing ECPR published from January 2000 to September 2019 was carried out in four English databases (PubMed, OVID, Embase, and ISI Web of Science). The following searching keywords were used, "extracorporeal membrane oxygenation OR extracorporeal oxygenation OR extracorporeal circulation OR extracorporeal cardiopulmonary resuscitation OR extracorporeal life support OR ECMO OR ECPR OR E-CPR OR ECLS", "out-ofhospital cardiac arrest OR in-hospital cardiac arrest OR cardiac arrest OR heart arrest OR OHCA OR IHCA", and "Hypothermia OR therapeutic hypothermia OR induced hypothermia OR mild hypothermia OR cooling OR cryotherapy OR targeted temperature management OR TTM". The retrieved articles were screened and selected according to the inclusion and exclusion criteria by two independent investigators. The references of the selected articles and the review articles were also screened to identify additional eligible studies.

Inclusion and exclusion criteria
Inclusion criteria, [1] subjects were adult patients with CA undergoing ECPR [2]; the clinical outcomes reported in the studies were neurological outcomes (primary outcomes) and patients' survival (secondary outcomes) at hospital discharge/30 days/90 days; the neurological outcomes should be assessed by Cerebral Performance Categories (CPC) score and CPC 1 or 2 was defined as favorable outcomes while CPC 3-5 were defined as poor outcomes [38] [3]; the target temperature of therapeutic hypothermia was stated [4]; the clinical outcomes between therapeutic hypothermia intervention and controls were compared [5]; unadjusted or adjusted odds ratios (ORs) with their 95% confidence intervals reflecting the associations of therapeutic hypothermia with clinical outcomes were reported or the proportions of the patients with unfavorable and favorable outcomes in therapeutic hypothermia and control arms were reported [6]; the language was limited as English. Exclusion criteria, [1] studies included other disease besides CA and the results for CA could not be separated [2]; studies performed in neonates/children/pediatric patients [3]; studies included other treatments besides ECPR (such as CCPR) and results for ECPR could not be separated [4]; studies with insufficient data [5]; studies with sample size< 10 [6]; studies with duplicate data; the author information, study centers, population, and the results of the candidate studies were compared to identify the studies with fully or partly duplicate data; when studies with duplicate data were identified, the studies with larger sample size and more recently published were included [7]; all patients were treated with therapeutic hypothermia (i.e. no controls) [8]; the target temperature of therapeutic hypothermia was not reported [9]; not original full articles, such as abstracts, letters, or review articles; and [10] case series or case report.

Data extraction
The following information was collected by two authors independently at the same time from each eligible study, the author information, publication year, countries, study period, gender, age, sample size, location of CA, bystander CPR attempt, rhythm, follow-up points, outcomes, indication criteria of ECPR and therapeutic hypothermia, target temperature of therapeutic hypothermia and control arms, the comparability of baseline characteristics, CA characteristics and neurological status before therapeutic hypothermia treatment, patients number with unfavorable/favorable outcomes in therapeutic hypothermia and control arms or direct ORs with 95%CIs, and analysis method (univariate or multivariate analysis). We also tried to contact with the authors of the included studies for detailed results needed in our study. Disagreement was solved by discussion and if the disagreement was still not solved, the arbitration would be performed by a third author.

Quality score assessment
The methodological quality of each included article was evaluated by a modified Newcastle Ottawa Scale (mNOS) with scores ranging from 0 to 9 (Additional file 1: Table S1). mNOS score ≥ 7 was considered as high quality.

Statistical analysis
The odds ratios (ORs) with 95% CIs from each eligible study were pooled to assess the strength of the associations of therapeutic hypothermia with neurological outcomes or survival in adult CA patients undergoing ECPR. If 30 days endpoints were not reported, discharge endpoints were used. The longest available follow-up endpoints were used in overall analysis and all follow-up endpoints were used when performing subgroup analysis according to follow-up period. The association was considered significant if P < 0.05 by Z-test. Chi-square based on Q statistic test was used to quantify the heterogeneity in each analysis. If P>0.1/I 2 <50%, ORs were pooled with fixed-effects model (Mantel-Haenszel method); otherwise, the random-effects model (DerSimonian-Laird method) was used [39]. Begg's funnel plots and Egger's test were used to determine the potential publication bias [40,41]. The Duval and Tweedie's trimand-fill method, a funnel plot-derived and two-step method, was used to estimate the effect size accounting for publication bias. The stability of the pooling result was verified by sensitivity analysis through omitting each study in each turn. In addition, subgroup analyses according to control type, regions, sample size, CA location, ORs obtained methods, follow-up period, and mNOS scores were also performed. Stata 12.0 software (Stata Corp, College Station, TX, USA) was used to perform all the analyses. Figure 1 showed the literature searching and screening process. According to the searching strategy, a total of 3792 documents were initially retrieved from four English databases (PubMed, OVID, Embase and ISI Web of Science) and additional sources. After excluding the duplicated papers, review or meeting abstracts, and irrelevant articles, 60 articles were left for further screening by checking the full texts. 47 articles were excluded, of which, seven were overlapped studies [42][43][44][45][46][47][48], seven were performed in pediatric patients [49][50][51][52][53][54][55], two study contained patients with other disease besides CA and the data related to CA could not be obtained separately [16,56], three studies contained CA patients treated with other methods besides ECPR and the data related to ECPR could not be obtained separately [17,57,58], fifteen studies had no sufficient data [59][60][61][62][63][64][65][66][67][68][69][70][71][72][73], and eleven studies only had therapeutic hypothermia arms (without controls) [37,[74][75][76][77][78][79][80][81][82][83]. Two studies did not state the specific temperature of therapeutic hypothermia and control arms [84,85]. Finally, 13 papers containing 1159 cases were included in the present study [9,[25][26][27][28][29][30][31][32][33][34][35][36]. Notably, two papers published by Kagawa et al. in 2012 and 2015 with some overlapped participants were both included because they reported different clinical outcomes [25,27]. Studies by Kim

Study characteristics
The detailed characteristics of the included 13 articles were shown in Table 1 and Additional file 2: Table S2. The studies were published from 2012 to 2019 and the samples size ranged from 10 to 274. The studies were conducted in Korea (n = 5), Japan (n = 6), Australia (n = 1), and Singapore (n = 1). The proportion of patients receiving therapeutic hypothermia, with favorable neurological outcomes, and with survival ranged from 13.14 to 78.26%, 13.92 to 35.14%, and 14.00 to 43.48%, respectively. Seven studies enrolled only OHCA patients and six studies enrolled both IHCA and OHCA patients. Nine studies reported the association of therapeutic hypothermia with neurological outcomes in adults CA patients undergoing ECPR [9, 26-28, 30-32, 34, 36], and nine reported the association of therapeutic hypothermia with patients' survival [9, 25, 28-31, 33, 35, 36]. Quality assessment was performed according to a modified NOS (Additional file 1: Table S1). The mNOS scores ranged from 6 to 9 for neurological outcomes (Additional file 1: Table S3) and from 6 to 8 for survival outcomes (Additional file 1: Table S4).

Meta-analysis results
Association of therapeutic hypothermia with neurological outcomes in CA patients receiving ECPR There were nine studies with a total of 806 cases reporting the association of therapeutic hypothermia with neurological outcomes in CA patients receiving ECPR. Six studies reporting neurological outcomes at hospital discharge and three studies reporting that at 90 days. Pooling analysis suggested that therapeutic hypothermia was significantly associated with favorable neurological outcomes in CA patients receiving ECPR in overall (N = 9, OR = 3.507, 95%CI = 2.194-5.607, P < 0.001, fixedeffects model) and in all subgroups according to control type, regions, sample size, CA location, ORs obtained   methods, follow-up period, and mNOS scores ( Fig. 2a and Table 2).

Association of therapeutic hypothermia with survival in CA patients receiving ECPR
There were nine studies with a total of 806 cases assessing the association of therapeutic hypothermia with survival in CA patients receiving ECPR. After pooling the ORs, therapeutic hypothermia was found to be significantly associated with favorable survival in overall (N = 9, OR = 2.540, 95%CI = 1.245-5.180, P = 0.010, random-effects model; Fig. 2b and Table 3). After subgroup analysis according to control type, regions, sample size, CA location, ORs obtained methods, follow-up

Heterogeneity, sensitivity analysis and publication bias
No heterogeneity existed when assessing the association of therapeutic hypothermia with neurological outcome while heterogeneity was found when evaluating the association of therapeutic hypothermia with patients' survival.  Fig. 3a and b). Begg's and Egger's tests were used to assess the publication bias. Publication bias was found when evaluating the association of therapeutic hypothermia with neurological outcomes (Fig. 4a and Table 2; P begg = 0.048, P egger = 0.008). Trim-and-fill analysis suggested that there were four "missing" studies on the bottom side of the funnel plot (Fig. 4b). After adjustment, therapeutic hypothermia was still significantly associated with favorable neurological outcomes in CA patients receiving ECPR (OR = 2.800, 95%CI = 1.842-4.526, P < 0.001, fixed-effects model). Conversely, no publication bias was found when evaluating the association of therapeutic hypothermia with patients' survival ( Fig. 4c and Table 3; P begg = 0.754, P egger = 0.703).

Discussion
Therapeutic hypothermia can decrease cerebral metabolism, alleviate the ischemic loss of ion gradients, reduce free radicals, and decrease postischemic inflammation after cerebral ischemia, thus protect patients against hypoxic brain damage following CA [86,87]. Therapeutic hypothermia has been recommended for eligible CA patients in order to improve outcomes and is increasingly used [22][23][24]. However, whether it improves neurological outcomes and survival in ECPR treated CA patients is not conclusive and the results in recent research are not consistent. In addition, therapeutic hypothermia can bring significant major complications including coagulation dysfunction, infection, and other systemic problems [88]. This promoted us, for the first time, to perform a systematic review with meta-analysis to summarize the effects of therapeutic hypothermia in CA patients receiving ECPR, and we found therapeutic hypothermia might be associated with favorable neurological outcomes and survival in CA patients receiving ECPR in overall. Therapeutic hypothermia is intended to be performed in CA patients with poor neurological status. Of the 13 included studies, seven studies describe the indications of therapeutic hypothermia treatments [9,25,27,29,31,34,36]. All the unconscious patients after ECPR are recommended receiving therapeutic hypothermia in six studies and one study [31] randomly divides the unconscious CA patients into therapeutic hypothermia group and non-therapeutic hypothermia group. In other words, the patients in therapeutic hypothermia arms usually have a poorer neurological status at baseline than those in control arms, which might act as a bias when evaluating the associations of therapeutic hypothermia with patients' outcomes. We pooled the seven studies and found that therapeutic hypothermia was associated with patients' favorable neurological outcomes (N = 4, OR = 3.627, 95%CI = 1.533-8.468, P = 0.003, fixed-effects model) and survival (N = 4, OR = 3.248, 95%CI = 1.909-5.525, P < 0.001, fixedeffects model), indicating that although the patients in therapeutic hypothermia arm might have a poorer baseline neurological status, an improved prognosis was still identified.
In 2002, two randomized clinical trials (RCTs) suggest that therapeutic hypothermia at 33°C improves patients' neurological outcomes compared to any temperature controls in CA [19,89]. Recently, Nielsen et al. perform a large RCT and find no differences of patients' neurological outcomes and survival between target temperature managements at 33°C and 36°C [90]. However, it is difficult to extrapolate this result to the ECPR population. In 2015, Kagawa et al. perform an observational study on 237 patients receiving targeted temperature management after CA and find no difference of neurological outcomes between targeted temperatures management at < 34°C and at ≥34°C in entire CA patients [27]. Nevertheless, they find < 34°C is associated with improved neurological outcomes in combined ECPR treated CA patients. In the present study, we observed improved neurological outcomes after therapeutic hypothermia induction at ≤34°C compared to no therapeutic hypothermia induction (N = 4, OR = 4.794, 95%CI = 2.037-11.28, P < 0.001, fixed-effects model) or alternative targeted temperature controls (> 34°C) (N = 5, OR = 3.067, 95%CI = 1.750-5.374, P < 0.001, fixed-effects model), respectively, in CA patients undergoing ECPR. Therapeutic hypothermia is associated with OHCA patients compared with IHCA patients in practice [61]. In the 13 included studies, the proportion of patients receiving therapeutic hypothermia ranges from 20.74 to 78.26% with an average of 41.13% in studies enrolling  Selection criteria for starting ECPR and characteristics of CA including cause of CA may also affect our results stability. Almost of our enrolled studies have stated the ECPR inclusion and exclusion criteria (Additional file 2: Table S2), however, these criteria are diverse, and we can't solve this problem based on current data. Characteristic of CA, such as the proportion of the patients with IHCA, witnessed CA, bystander CPR attempt, shockable rhythm, cardiac origin CA, or acute coronary syndrome, are also different in our enrolled studies (Additional file 2: Table S2). We have used a metaregression method to explore the effects of the characteristic of CA on the association of therapeutic hypothermia with neurological outcomes in CA patients receiving ECPR and find these characteristics have no effect on their association (Additional file 1: Table S5).
There are other shortcomings that may impact our results stability. Firstly, the eligible studies and sample size are relative limited although we try our best to search all the potential eligible studies. Secondly, all the eligible studies are retrospective observational studies and most of them do not compare the baseline characteristics, CA characteristics, and initial neurological status before therapeutic hypothermia treatment. Thirdly, most of the included studies have not reported adjusted ORs. Finally, the eligible studies are conducted in only several regions including Korea, Japan, Singapore, and Australia. Thus, further prospective well-designed and multicentric random controlled studies with larger sample size should be conducted to validate our conclusions.

Conclusions
We got a comprehensive result from the current metaanalysis that therapeutic hypothermia might be associated with favorable neurological outcomes and survival in adult patients undergoing ECPR after cardiac arrest. However, our result is a synthesis of retrospective observational studies, which are the low-level evidence, and should be treated carefully. Further prospective welldesigned and multicentric randomized controlled trials with larger sample size are necessary before putting our result into clinical practices.
Additional file 1: Table S1. Modified Newcastle-Ottawa scale (mNOS) used in present study for cohort studies. Table S3. Modified Newcastle-Ottawa scale (mNOS) scores of our included studies for neurological outcomes. Table S4. Modified Newcastle-Ottawa scale (mNOS) scores of our included studies for patients' survival. Table S5. Meta-regression analyses on characteristics of cardiac arrest for the associations of therapeutic hypothermia with favorable neurological outcomes in CA patients receiving ECPR.
Additional file 2. Detailed characteristics of the included studies.