Mechanical versus manual chest compressions for out-of-hospital cardiac arrest: a meta-analysis of randomized controlled trials

Recent evidence regarding mechanical chest compressions in out-of-hospital cardiac arrest (OHCA) is conflicting. The objective of this study was to perform a meta-analysis of randomized controlled trials (RCTs) to compare the effect of mechanical versus manual chest compressions on resuscitation outcomes in OHCA. PubMed, Embase, the Cochrane Central Register of Controlled Trials, and the ClinicalTrials.gov registry were searched. In total, five RCTs with 12,510 participants were included. Compared with manual chest compressions, mechanical chest compressions did not significantly improve survival with good neurological outcome to hospital discharge (relative risks (RR) 0.80, 95% CI 0.61–1.04, P = 0.10; I2 = 65%), return of spontaneous circulation (RR 1.02, 95% CI 0.95–1.09, P = 0.59; I2 = 0%), or long-term (≥6 months) survival (RR 0.96, 95% CI 0.79–1.16, P = 0.65; I2 = 16%). In addition, mechanical chest compressions were associated with worse survival to hospital admission (RR 0.94, 95% CI 0.89–1.00, P = 0.04; I2 = 0%) and to hospital discharge (RR 0.88, 95% CI 0.78–0.99, P = 0.03; I2 = 0%). Based on the current evidence, widespread use of mechanical devices for chest compressions in OHCA cannot be recommended.

Data extraction and outcomes definition. Two authors (Tang L and Gu WJ) independently extracted the following data: first author, year of publication, study location, participant characteristics, mechanical compression device, resuscitation strategy, adverse events, and main outcomes using a standard form. The original authors were contacted if data needed clarification or were not presented in the publication. Any disagreement was resolved by discussion.
The primary outcome was survival with good neurological outcome to hospital discharge. Secondary outcomes included survival to hospital admission, survival to hospital discharge, ROSC and long-term (≥ 6 months) survival. Neurological outcome was assessed using the Cerebral Performance Category (CPC) score or the modified Rankin Scale (mRS) score. Good neurological outcome was defined as the CPC score ≤ 2 14 or the mRS score ≤ 3 15 . Survival to hospital admission was defined as patient admission to the hospital without ongoing CPR or other artificial circulatory support. ROSC was defined as the presence of any palpable pulse, in the absence of chest compression, and detectable by manual palpation. When data on survival to hospital admission was not available, we extract data on 4-hour survival after successful ROSC. In case that survival to hospital discharge was not reported, data on survival to 30 days were extracted. With respect to long-term survival, data on survival to six months or more were extracted.
Risk of bias and evidence grade assessment. Two authors (Tang L and Gu WJ) independently assessed risk of bias in included RCTs with the method recommended by the Cochrane Collaboration 16 . Disagreements were resolved by consensus. The quality of evidence for the outcome measures was evaluated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach 17 . A summary table was prepared using the GRADE profiler (GRADEpro, version 3.6).
Statistical analysis. All analyses were on an intention-to-treat basis. Differences were presented as relative risks (RRs) with 95% confidence intervals (CIs) for dichotomous outcomes. The Mantel-Haenszel method with the random-effects model was used to calculate pooled RRs and 95% CIs. Heterogeneity was quantified using the I 2 statistic and if the I 2 value was greater than 50%, heterogeneity was considered to be substantial. A priori subgroup analysis was conducted according to mechanical compression devices (LDB-CPR and PD-CPR). Publication bias was assessed by visually inspecting a funnel plot. P < 0.05 was considered statistically significant. All statistical analyses were conducted using Review Manager software (version 5.3; Nordic Cochrane Centre, Cochrane Collaboration).

Results
Study identification. The comprehensive search yielded 648 citations, of which 633 were eliminated for various reasons based on the title and abstract. The full texts of the remaining 15 publications were scrutinized for further evaluation. Finally, five RCTs 9-11,18,19 with a total of 12,510 participants were included (Fig. 1). The Cohen statistic К for agreement on study inclusion was 0.93.

Study characteristics.
The main characteristics of the included RCTs are presented in Table 1. The trials were published between 2006 and 2015 and were all multicenter studies. The sample size ranged from 148 to 4471. Among the included trials, two 10,18 compared LDB-CPR with manual CPR and the remaining three compared PD-CPR with manual CPR. Four 9,10,18,19 included trials were sponsored by the manufacturer of the mechanical compression devices, and the remaining one trial 11  Risk of bias assessment. Randomized sequence was adequately generated in two trials 10,11 and allocation sequence concealment was adequately reported in four trials [9][10][11]19 . Although blinding of participants and personnel and blinding of outcome assessors were difficult to implement due to the nature of the intervention, survival and neurological outcome assessment were very unlikely to have been influenced by knowledge of trial allocations. Thus, the two items (blinding of participants and personnel and blinding of outcome assessors) were judged to be low risk of bias for all the included trials. The numbers and reasons for withdrawal/dropout were detailed reported in all trials. One trial 18 was defined as having other sources of bias because it was stopped early after the interim analysis. An overview of the risk of bias is summarized in Fig. 2.
Primary outcome. Data on the primary outcome was available in four of the included RCTs (n = 12,058). Mechanical chest compressions did not significantly improve survival with good neurological outcome to hospital discharge (RR 0.80, 95% CI 0.61-1.04, P = 0.10; I 2 = 65%; Fig. 3) compared with manual chest compressions. Subgroup analyses stratified by mechanical compression devices also suggested no significant difference in the primary outcome both in the LDB-CPR subgroup and in the PD-CPR subgroup.
Quality of evidence and publication bias. The GRADE evidence profiles for the primary and secondary outcomes were shown in Table 2. The quality of evidence was low for survival with good neurological outcome to hospital discharge; moderate for survival to hospital admission, survival to hospital discharge, ROSC and long-term (≥ 6 months) survival.
For publication bias, the funnel plot was not conducted due to the small number of RCTs included.

Discussion
The accumulated evidence from meta-analysis of RCTs suggested that mechanical chest compressions did not improve survival with good neurological outcome to hospital discharge, ROSC or long-term (≥ 6 months) survival, and were associated with worse survival to hospital admission and to hospital discharge in OHCA compared with manual chest compressions. Differences between the current meta-analysis and a previous one by Westfall et al. 8 should be noted. First, Westfall et al. 8 conducted a meta-analysis of English-language randomized and non-randomized (phased, historical, and case-control) studies. Thereinto, only 2 small studies were RCTs 18,19 and most of the included studies were non-randomized, thereby producing a potential for selection bias. Our meta-analysis included only RCTs and three large RCTs 9-11 published since 2014 were included, which substantially enlarged the sample size (12,510 vs. 6,538 participants). Second, considering that survival data are deemed to be one of the highest quality and/or most clinically relevant endpoints for a given trial, we chose survival with good neurological outcome to hospital discharge, a more patient-centered outcome measure, as the primary outcome. Additionally, survival to hospital admission, survival to hospital discharge, ROSC and long-term (≥ 6 months) survival were also reported in the present meta-analysis. However, in Westfall et al.'s meta-analysis 8 , the primary outcome was ROSC and survival data were not reported. Third, the methods used for risk of bias assessment and GRADE profile evidence were described in our meta-analysis not in Westfall  In the present meta-analysis, the findings of worse survival to hospital admission and to hospital discharge in participants receiving mechanical chest compressions were unexpected. One possible explanation is that interruptions in CPR during device deployment could cause reduced cardiac and cerebral perfusion. Furthermore, device deployment before the first defibrillation is likely to have led to a delay in the time to first defibrillation, which might in itself reduce survival. In one included RCT 18 , the time to first defibrillation was delivered 2.1 minutes later in the LDB-CPR group than in the manual CPR group. Similarly, in one included RCT evaluating the PD-CPR device 9 , the mechanical compression group had an average 1.5-minute extra delay to first defibrillation attempt. Therefore, if mechanical chest compression devices are being used in OHCA, special attention should be paid to minimizing the delay to chest compressions and defibrillation related to deployment of the device.
Since interruptions in CPR and delays in device deployment are a major factor that can impact outcomes, intensive and repetitive training is particularly important when using the mechanical compression devices in OHCA. In Perkins et al.'s study 11 , only 60% participants in the PD-CPR group received the allocated intervention and about 15% cases of non-use were because of difficulties inherent with implementation of new equipment and the training and quality issues associated with this. In a recent study 20 , mechanical chest compressions were associated with a higher no-flow ratio than manual chest compressions in the first five minutes of resuscitation. Where organizations decide to adopt mechanical  Table 1. Characteristics of included randomized controlled trials. Data are presented as mean ± SD unless indicated otherwise. UK, united kingdom; US, united states; CPR, cardiopulmonary resuscitation; PD-CPR, piston-driven CPR; LDB-CPR, load-distributing band CPR; M-CPR, manual CPR; EMS, emergency medical systems; VF, ventricular fibrillation; VT, ventricular tachycardia. * Mean (range). compression device, it seems essential that sufficient resources are made available to support initial and regular refresher training and ongoing quality assurance.
Clinical adverse events occurring in chest compressions were reported in three included RCTs 9-11 , such as rib fractures, pulmonary edema, pneumothorax and chest bruising. Specifically, the type of device and the mechanism of the device were thought to be of paramount relevance to injury patterns observed in OHCA. A non-randomized study reported that the PD-CPR device slid from its original position, mostly in an abdominal direction during transport 21 . Although pooled analyses were not  conducted due to significant clinical heterogeneity, no significant difference in serious adverse events was reported in individual included RCT.
The quality of evidence in the present meta-analysis ranked from moderate to low across the different outcomes. The limiting factor that was the reason for a decrease in quality for all outcomes was the serious risk of bias. Only two of the included RCTs were judged to be at low risk of bias. Additionally, due to the unexplained heterogeneity (I 2 = 65%) across the small number of included studies, the evidence   quality for the survival with good neurological outcome to hospital discharge was downgraded by one more level and was therefore assessed as low quality.
There are several limitations in the present study. First, although comprehensive literature search was conducted, the number of included RCTs was small and the funnel plot for publication bias was not conducted. Second, although more and more services emphasize the importance of ongoing performance monitoring for out-of-hospital CPR quality, none of the included RCTs reported any measures of CPR quality and we cannot assess the influence of the quality of manual CPR on the treatment effect estimates. Future studies comparing mechanical with manual CPR should include the instrumentation to monitor CPR quality. Third, although subgroup analyses suggested that LDB-CPR not PD-CPR significantly decreased survival to hospital admission and to hospital discharge than manual CPR, caution should be taken when interpreting the results because they were based on the pooled analysis of only two RCTs.  Table 2. GRADE evidence profile. 1 Only two trials were judged to be at low risk of bias. 2 Substantial heterogeneity (I 2 = 65%) was found. 3 Only one trial was judged to be at low risk of bias.

Conclusions
Our meta-analysis suggested that mechanical chest compressions were not associated with better outcomes including survival with good neurological outcome to hospital discharge, survival to hospital admission, survival to hospital discharge, ROSC and long-term (≥ 6 months) survival in OHCA compared with manual chest compressions. Based on the current evidence, widespread use of mechanical devices for chest compressions in OHCA cannot be recommended.