Effects of prehospital adrenaline administration on out-of-hospital cardiac arrest outcomes: a systematic review and meta-analysis

Introduction The aim of this study was to conduct a systematic review and meta-analysis for determining the effects of prehospital adrenaline administration on return of spontaneous circulation, hospital admission, survival to discharge and discharge with cerebral performance category 1 or 2 in out-of-hospital cardiac arrest patients. Methods MEDLINE and Scopus databases were searched to identify studies reported to March 2014. Study selection and data extraction were independently completed by two reviewers (PA and SR). The baseline characteristics of each study and number of events were extracted. Risk ratios (RR) and 95% confidence interval (CI) were estimated. Heterogeneity and publication bias were also explored. Results In total 15 studies were eligible and included in the study. Of 13 adult observational studies, four to eight studies were pooled for each outcome. These yielded a total sample size that ranged from 2,381 to 421,459. A random effects model suggested that patients receiving prehospital adrenaline were 2.89 times (95% CI: 2.36, 3.54) more likely to achieve prehospital return of spontaneous circulation than those not administered adrenaline. However, there were no significant effects on overall return of spontaneous circulation (RR = 0.93, 95% CI: 0.5, 1.74), admission (RR = 1.05, 95% CI: 0.80, 1.38) and survival to discharge (RR = 0.69, 95% CI: 0.48, 1.00). Conclusions Prehospital adrenaline administration may increase prehospital return of spontaneous circulation, but it does not improve overall rates of return of spontaneous circulation, hospital admission and survival to discharge. Electronic supplementary material The online version of this article (doi:10.1186/s13054-014-0463-7) contains supplementary material, which is available to authorized users.

Although three systematic reviews [22][23][24] have been conducted, a more complete review was still required for the following reasons. One previous review studied a mix of in-and out-of hospital cardiac arrest patients without providing any quantitative conclusion [22], and another systematic review pooled adrenaline effect on survival to hospital discharge but not for prehospital ROSC [23].
Finally, Lin et al. conducted a systematic review by including only randomized controlled trials (RCTs), which primarily aimed to compare the standard dosages of adrenaline administration with placebo, vasopressin and high dosage of adrenaline in out of hospital cardiac arrest patients [24]. Although those authors found benefit of adrenaline administration over placebo on ROSC and survival to admission, the results were based on only one RCT. There have been more studies published since these reviews were reported. We, therefore, conducted a systematic review and meta-analysis focused only on out of hospital cardiac arrest patients, which aimed to determine the effects of prehospital adrenaline administration on both short (that is, ROSC, hospital admission) and long term outcomes (that is, survival to discharge and discharged with cerebral performance category (CPC) 1 or 2). Overall ROSC = 4

Search strategy
This study followed Meta-Analysis Of Observational Studies in Epidemiology (MOOSE) guidelines [25]. Because we worked on public data, neither approval nor patient consent was required by the Ethics Committee of the Faculty of Medicine, Ramathibodi Hospital, Mahidol University. Two reviewers (PA and SR) identified studies from MEDLINE and Scopus databases since conception to March 2014 using PubMed and Scopus search engines. The following search terms were used: (heart arrest or out of hospital cardiac arrest or ventricular fibrillation or pulseless electrical activity or PEA or asystole or cardiac arrest) and (epinephrine or adrenaline) and (return of spontaneous circulation or ROSC or admission or discharge or death or survival). Studies were firstly selected based on title and abstracts, and full papers were reviewed to make a final decision on selection. Any type of study (that is, RCT, quasi-experimental study, cohort, or cross sectional study) published in English was selected if it met the following eligibility criteria: studied in patients with out of hospital cardiac arrest, compared clinical outcomes between prehospital adrenaline and no administration, and had at least one of these clinical outcomes of interest: ROSC, hospital admission, survival to hospital discharge, or discharged with CPC 1 or 2. Studies were excluded if there were insufficient data for pooling and if authors did not provide additional data after being contacted twice.
The short-term outcomes of interest included ROSC and hospital admission and the long-term outcomes of interest were survival to discharge and discharged with CPC 1 or 2. The ROSC was defined according to individual original studies as any sign of spontaneous circulation including palpable pulse or measurable blood pressure following CPR [13,16,26,27] occurring during prehospital, at emergency department (ED) arrival, or in-hospital. Survival to discharge referred to patients who were still alive at discharge from hospital. Neurological outcome at discharge was assessed using the Glasgow-Pittsburgh cerebral performance category scale, which categorizes patients as CPC 1 to 5 as follows: good performance, moderate, severe, coma/vegetative, and death, respectively [14,18,[26][27][28].

Data extraction
Data were extracted by two independent reviewers (PA and SR). Study characteristics (that is, settings, study designs, types of subjects, mean age and percentage of males) were extracted. In addition, clinical data including initial cardiac rhythms, dose and routes of adrenaline administration, presumed cardiac etiologies, witnessed cardiac arrest and bystander CPR status were also extracted. The numbers of subjects who did and did not experience outcomes of interest among adrenaline administration groups were also extracted. Any disagreement was discussed with the senior author (AT).

Risk of bias assessment
The Newcastle-Ottawa Scale for assessing the quality of a cohort study [29] was modified and used to assess risk of bias for included cohort studies (see Additional file 1: Table  S2). This tool consists of five items, which are representative of out of hospital cardiac arrest cohorts, ascertainment of exposure and outcomes, adjusting for confounders and missing data. Each item was graded as low or high risk of bias, and unclear if insufficient information.

Statistical analysis
The effect of adrenaline administration was estimated for each study using risk ratio (RR) along with its 95% confidence interval (CI). The RR was pooled across studies using the random-effect model if heterogeneity was present [30], otherwise a fixed-effects model was applied. The criteria for declaring heterogeneity was a P-value of the Cochrane's Q statistics <0.1 or I 2 25% or higher [31]. Sources of heterogeneity were explored using Galbraith's plot and meta-regression analysis. Each potential variable (that is, types of subjects, initial cardiac rhythms and proportion of witnessed cardiac arrest) was fitted in   a meta-regression model. A sensitivity analysis was applied by excluding candidate studies suspected to be a source of heterogeneity. Publication bias was assessed using a funnel plot and Harbord's test [32]. A contour-enhanced funnel plot was applied to explore whether there was any cause of asymmetry due to publication bias or heterogeneity [33]. All analyses were performed using STATA version 12.0. A P-value of less than 0.05 was considered as statistically significant, except for the heterogeneity test in which 0.10 was used.

Risk of bias assessment
Thirteen adult cohort studies were assessed for risk of bias using the Newcastle-Ottawa scale (see Additional file 1: Table S3). Eleven studies were low risk of bias for representativeness of out of hospital cardiac arrest. There were 11 studies which were graded as low risk of bias for both ascertainment of adrenaline administration and outcomes. Nine studies assessed treatment effects by adjusting confounders, thus they were low risk of confounding bias. Finally, 10 studies were graded as low risk of bias for missing data.

Return of spontaneous circulation Prehospital ROSC
Six adult cohorts [12][13][14][15]35,36] were included for pooling adrenaline administration effects on prehospital ROSC with sample sizes of 16,321 for adrenaline and 405,138 for non-adrenaline groups, see Table 2. The prehospital adrenaline effect was highly heterogeneous across studies (Q = 153.2, d.f. = 5, P <0.001 , I 2 = 96.7%), see Figure 2A and Table 2. A random-effects model was applied and yielded a pooled RR of 2.89 (95% CI: 2.36, 3.54), indicating that a patient receiving prehospital adrenaline was 2.89 times more likely to experience prehospital ROSC than one not receiving prehospital adrenaline.
A funnel plot showed a departure from symmetry of the funnel (see Figure 2B), but this was not detected by the Harbord's test (coefficient = −7.92, P = 0.134). However, a contour-enhanced funnel plot was constructed to explore any cause of asymmetry (see Additional file 1: Figure S1). This suggested that all studies fell in the significant area indicating that asymmetry of the funnel was more likely due to missing studies (that is, publication bias) than heterogeneity.

Overall return of spontaneous circulation
The effect of prehospital adrenaline on overall ROSC was pooled based on four studies with a total sample  Table 3 and Figure 3A. This suggested no effect of prehospital adrenaline on achieving overall ROSC. No source of heterogeneity could be identified from Galbraith's plot or meta-regression analysis. There was evidence of asymmetry from the funnel plot and the Harbord's test (coefficient = −9.27, P = 0.005), see Figure 3B. The contour enhanced funnel plot showed that 50% of the studies fell in both significant and non-significant areas, indicating heterogeneity was most likely the cause of asymmetry of the funnel (see Additional file 1: Figure S2).
The funnel plot showed a little asymmetry of the funnel, which corresponded with the Harbord's test (coefficient = 0.03, P = 0.988), see Figure 4B. A contourenhanced funnel plot showed that most studies fell in the non-significant area, suggesting that asymmetry was more likely due to heterogeneity (see Additional file 1: Figure S3).

Survival to discharge
Data of seven cohorts [5,14,16,[18][19][20]35] were used for pooling effects of prehospital adrenaline administration on survival to discharge, see Table 5. The pooled RR was 0.69 (95% CI: 0.48, 1.00) with a moderate degree of heterogeneity (Q = 9.1, d.f. = 6, P = 0.049, I 2 = 34.1%), see Table 5 and Figure 5A. This suggested that receiving prehospital adrenaline yielded about a 31% lower chance of discharge alive, but this was of borderline significance. There was no evidence of asymmetry of the funnel as suggested by the Harbord's test (coefficient = −0.37, P = 0.697) and the funnel plot, see Figure 5B.

Discussion
We performed a systematic review and meta-analysis to assess the effects of prehospital administration of adrenaline on short-term (that is, ROSC and hospital admission) and long term out of hospital cardiac arrest outcomes (that is, survival to discharge and discharge with CPC 1 or 2) relative to no administration. Pooled treatment effects were performed based on four to eight adult cohorts with total sample sizes of 2,381 to 421,459. Our results indicated that receiving prehospital adrenaline significantly increased prehospital ROSC, but did not increase the overall ROSC, hospital admission, or survival to hospital discharge. However, the treatment effects on prehospital ROSC may be prone to bias from missing studies.
Administration of adrenaline during CPR causes constriction of peripheral vessels, which subsequently increases coronary and cerebral perfusion pressure [3]. Our findings support the effect of adrenaline in increasing prehospital ROSC, which is similar to a previous RCT [27] and systematic reviews by Lin et al. [24]. However, our effect size was more precise than the finding by Lin et al. because the pooling was based on a larger number of included studies than Lin et al., which was based on only one study. In addition, our findings also suggested the adrenaline effect on ROSC in both nontrauma and the mix of trauma and non-trauma cases, although the latter was highly heterogeneous. However, there was no adrenaline effect on overall ROSC, which might be due to variation in the definition and time of ROSC assessment, that is, prehospital, ED, and admission [12,14,18,19,28,36]. Assessing ROSC at the ED or on admission could be very different compared to prehospital ROSC because of the quality of CPR, type of facility, and personnel and equipment used.
No other long term effects (that is, hospital admission and survival to discharge) of prehospital adrenaline were identified. This suggests that adrenaline itself may only have efficacy for inducing prehospital ROSC. Ultimately, survival to hospital discharge is determined more by the underlying clinical condition of the patient; for example, a patient with acute myocardial infarction may be more likely to survive than a patient with end stage respiratory disease, and yet most studies [5,14,18] did not report the underlying diagnoses which leads to difficulties in interpreting the results. In addition, adrenaline is intrinsically a short acting cardiovascular stimulant, which has a limited half-life, and it may be less likely to have a significant effect on long term outcomes for this reason [3]. Assessing treatment effects without adjusting for these factors, as was the case for the included studies, may bias the results. Our study has a number of strengths. We pooled treatment effects on most relevant clinical short-and longterm outcomes. We attempted to identify the most relevant studies by two independent reviewers using defined search terms and strategies in order to reduce selection bias. Exhaustive searching of the literature resulted in the number of included studies from four to eight cohorts. Although the number of included studies is still small, but, it is larger than a previous systematic review [22], which included five studies [5,16,27,38,39] with different study designs. We explored possible sources of heterogeneity by considering whether a bystander witnessed the arrest, types of subjects and initial cardiac rhythms in a metaregression. In addition, an attempt was made to distinguish the cause of asymmetry of funnel plots to determine if this was due to heterogeneity or missing studies.
However, our study has some weaknesses. Firstly, our results might be subject to some bias because pooled effects were based on only observational studies, which could not adjust for several known and unknown confounders. The most suitable study design is a RCT with a proper sample size, say approximately 4,500 subjects in total to detect a difference of prehospital ROSC rate of 2.5. Secondly, a subgroup analysis could not be flexibly done because of the small number of included studies. Thirdly, insufficient information on study characteristics resulted in limited exploration of sources of heterogeneity. Finally, we did not pool more relevant clinical outcomes (that is, sustained ROSC and discharge with CPR 1 or 2) because data were not available.

Conclusions
In summary, prehospital adrenaline administration might increase prehospital ROSC, but not for survival to discharge in out of hospital cardiac arrest patients. However, our findings might be subject to bias from missing studies.

Key messages
Currently, there was only one RCT which compared prehospital adrenaline administration versus placebo among out of hospital cardiac arrest patients. Receiving prehospital adrenaline significantly increased prehospital ROSC, but not for overall ROSC, hospital admission, and survival to discharge.

Additional file
Additional file 1: This file contains Table S1. Search strategy, Table S2. Modified Newcastle-Ottawa quality assessment scale for cohort studies, Table S3. Risk of bias assessment for cohort study, Figure S1. Contour-enhanced funnel plot of prehospital adrenaline effect on prehospital ROSC, Figure S2. Contour-enhanced funnel plot of prehospital adrenaline effect on overall ROSC, Figure S3. Contourenhanced funnel plot of prehospital adrenaline effect on hospital admission, Figure S4. Contour-enhanced funnel plot of prehospital adrenaline effect on survival to discharge.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions PA was responsible for the idea initiative, conception of the study design, study selection, data extraction, risk of bias assessment, data analysis, result interpretation and drafting the manuscript. SR participated in the conception of the study design, study selection, data extraction, risk of bias assessment, data analysis, result interpretation and drafting the manuscript. MM was involved in the conception of the study design, result interpretation, drafting the manuscript and critical review of the manuscript. CG participated in the conception of the study design, result interpretation, drafting the manuscript and critical review of the manuscript. YS contributed to the conception of the study design, result interpretation, drafting the manuscript and critical review of the manuscript. AT participated in the conception of the study design, risk of bias assessment, data analysis, result interpretation, drafting the manuscript and critical review of the manuscript. All authors read and approved the final manuscript.