Manual vs Device-Assisted CPR: Title and subTitle BreakReconciling Apparently Contradictory Results

Out-of-hospital sudden cardiac death is a major health problem. According to Becker et al,¹ in summarizing the 2000 Pulse Conference, “We lose more than 1000 lives each day in the United States from sudden, unexpected death, a fatality rate comparable to the crash of two 747 aircraft without survivors.” To make matters worse, current interventions for the treatment of nontraumatic cardiopulmonary arrest, conforming to a “chain of survival” concept, have not significantly improved neurologically intact survival rates over a decade.^{2 - 3} The overall survival rate is less than 5% and the chance of normal neurological function is even lower.

The considerations in resuscitation research can be conceptualized using a 3-phase model for resuscitation after cardiac arrest.⁴ In the first or electrical phase, immediate defibrillation of ventricular fibrillation, if present, is most likely to result in survival. To this end, the development of automated external defibrillators and their use by lay rescuers has been a major advancement. The results of the public access defibrillation trial have been published,⁵ the intervention accepted by the medical community, and automated external defibrillators can now be purchased at local drug stores.

In the second or circulatory phase, immediate countershock is unlikely to result in return of spontaneous circulation (ROSC), even if ventricular fibrillation is stopped. For patients with cardiac arrest rhythms other than ventricular fibrillation, this phase is actually the first phase. Two investigations^{6 - 7} have demonstrated that cardiopulmonary resuscitation (CPR) preceding countershock improves survival from cardiac arrest due to ventricular fibrillation. If defibrillation fails to restore spontaneous circulation, CPR, typically with pharmacotherapy, must be continued.

Several studies have shown that “good” CPR can save lives but that “good” CPR is difficult to perform for an extended period.^{8 - 9} Furthermore, frequent interruptions of chest compressions, occasioned by interposed artificial ventilations, pulse checks and rhythm analysis, can adversely impact outcome.¹⁰ Most individuals who sustain cardiac arrest probably receive suboptimal CPR, especially during extended resuscitation efforts.

During the third or metabolic phase of cardiac arrest, successful resuscitation becomes increasingly unlikely due to damage from global ischemia and the production of harmful metabolic factors with reperfusion.⁴ The phase during which an intervention is applied may profoundly influence the likelihood of benefit.

In an effort to improve manual CPR, a number of CPR devices have been developed and are in various stages of testing.¹¹ In this issue of JAMA, 2 studies^{12 - 13} are reported comparing the same load-distributing band (LDB) mechanical CPR device (LDB-CPR) with manual CPR for the treatment of out-of-hospital cardiopulmonary arrest of presumed cardiac etiology, yielding apparently contradictory results. Hallstrom and colleagues¹² conducted a prospective, cluster-randomized study in 5 communities, only to have the independent data and safety monitoring board recommend early termination because of lack of benefit in the primary outcome (survival to 4 hours) and apparent harm in secondary, longer-term, patient-centered outcomes. Ong and colleagues¹³ conducted a retrospective analysis of the incorporation of LDB-CPR in a single emergency medical service (EMS) system using a pre-post study design and found that LDB-CPR was associated with an improvement in the primary outcome, ROSC, and in multiple secondary, longer-term outcome measures.

The apparent contradiction in conclusions drawn from these 2 studies leads to several critical questions, including whether the true effectiveness of the device was different in the 2 settings, perhaps because of differences in patient populations or the details of the incorporation of the device into treatment of out-of-hospital cardiac arrest; whether some methodological weakness in the design, conduct, or analysis of 1 or both studies led to an incorrect or biased estimate of the effect of LDB-CPR; and whether anomalous outcome patterns in 1 or both trials occurred by chance (Table).^{12 - 13}

The study by Hallstrom et al¹² was a multicenter, cluster-randomized, prospective interventional trial, which because of the nature of the intervention could not be blinded. The multicenter design augments external validity, but its interpretation is complicated by the inclusion of 1 site (site C) that modified the intervention part way through the study and yielded systematically different (and better) outcomes than the other sites. Importantly, the evidence of harm from LDB-CPR in the secondary outcomes existed separately for patients at site C and those enrolled at the other 4 sites. The use of cluster randomization is used to reduce “contamination” of the treatment groups in health services trials, because individuals or teams who are providing care (eg, CPR) consistently apply only 1 treatment.^{14 - 15} As was performed by Hallstrom et al, the final statistical analysis must take the cluster randomization into account to avoid overestimating the statistical significance of observed differences.^{14 - 15}

Blood flow with the LDB-CPR device, as demonstrated in preclinical work, is largely the result of phased alterations in intrathoracic pressure, the so-called “thoracic pump” for CPR flow.¹⁶ It is unlikely that the poor outcome in the LDB-CPR group was related to inadequate blood flow. Cardiac output generated by a thoracic pump mechanism is adequate to sustain vital organ perfusion.

Standard, manual CPR may be better than is generally recognized. Preclinical investigations of the LDB-CPR device typically compared it with the only commercially available mechanical CPR device, a gas drive piston-cylinder. This device compresses the sternum 1.5 to 2 inches, presumably causing cardiac compression, the other mechanism of blood flow during CPR. In preclinical comparisons, the piston-cylinder device appears to have been adjusted to produce 20% anterior-posterior sternal displacement, similar to that produced by LDB-CPR and likely to be less than an inch in 20-kg swine.^{17 - 18} This 20% value may represent the optimal performance characteristics for LDB-CPR in small swine but may have reduced the performance of the older piston-cylinder device. A comparison of LDB-CPR to “good” manual CPR in a laboratory model could prove to be revealing.

Hallstrom et al¹² suggest a number of possible explanations for their failure to demonstrate improved outcome with the LDB-CPR device, including improved effectiveness of manual CPR due to the Hawthorne effect,¹⁹ the additional deployment time required for the LDB-CPR device, and enrollment bias resulting from EMS personnel “enthusiasm” for LDB-CPR. These alternative explanations are certainly plausible and more importantly could be corrected. Furthermore, in a post hoc analysis, there is some evidence of an interaction between response time and the effectiveness of LDB-CPR, in that patients with shorter response times appear to benefit from manual CPR and those with longer response times may benefit from LDB-CPR. This may be an important clue in understanding the results, if only patients in the later phases of cardiac arrest receive benefit from LDB-CPR.

The study by Ong et al¹³ used an observational pre-post study design which, although avoiding the implementation difficulty of the cluster-randomized design, is susceptible to bias from secular trends. The study consisted of 3 phases, a manual CPR phase, an intermediate phase during which the LDB-CPR device was introduced and during which no data collection occurred, and the LDB-CPR phase. This design has the advantage, however, that during each extended period the artificial nature and disruptions caused by randomization are avoided, and the use of the device in the EMS setting may most closely resemble routine clinical practice. The primary outcome was ROSC, with secondary outcomes of survival to hospital admission and hospital discharge, and neurological status. Improvements in ROSC, survival to hospital admission, and survival to hospital discharge were observed, even after adjustment for confounders.¹³

Under the usual hierarchy of experimental evidence from clinical studies, however, a multicenter, randomized, prospective study is believed to have greater external validity than a pre-post observational study from a single site. Moreover, even internal validity may be threatened by the influence that the introduction of a new CPR modality may have on other aspects of care for cardiac arrest. For example, the quality of CPR is an important determinant of outcome.⁸ It is plausible that the introduction of a mechanical CPR device would focus greater attention on the provision of good manual CPR even before the deployment of the device, as well as greater attention to avoiding pauses in chest compressions or ventilation. Thus, an improvement in clinical outcome after implementation may be a true effect of the introduction of the device that is, paradoxically, not related to the physiological effects of the device itself. Moreover, in checking for baseline differences between patients in the 2 treatment groups, it would be useful to consider those patients who never received LDB-CPR because of a rapid response to early resuscitative efforts (eg, defibrillation). These patients, who are excluded from the LDB-CPR group when the trial is analyzed by treatment received rather than by the intention-to-treat approach, are easy to identify in the LDB-CPR group but much more difficult to identify in the manual CPR group. This makes any comparison of the number of these easily resuscitated patients between the 2 treatment groups quite speculative.

To some extent, the most notable results of both these trials are contained in their secondary outcomes. In the study by Hallstrom et al, the apparent harm manifested in the secondary outcomes of survival to hospital discharge and neurological outcome prompted the data and safety monitoring board to recommend termination of the study; whereas, in the study by Ong et al, the difference in rates of ROSC would be relatively unimportant without a concomitant change in survival to hospital discharge and neurological outcome. Accordingly, perhaps the best clue regarding the contradictory conclusions is found in comparing the secondary outcomes between the 2 trials.

When considering survival to hospital discharge, it is noteworthy that manual CPR in the study by Hallstrom et al achieved the same result as LDB-CPR in the study by Ong et al—10% (Table). Similarly, the rates of a good neurological outcome (Cerebral Performance Category score of 1 or 2) are similar between the manual CPR group in the study by Hallstrom et al (8%) and the LDB-CPR group in the study by Ong et al (6%). With respect to this last outcome, one interpretation might be that manual CPR as administered in the study by Ong et al performed anomalously poorly (2% good neurological outcome); whereas, LDB-CPR as administered in the study by Hallstrom et al also performed anomalously poorly (3% good neurological outcome).

On April 5, 2006, the manufacturer of the AutoPulse LDB-CPR device published a press release in which they “reported that a growing number of EMS agencies in the State of Virginia are deploying the AutoPulse to help better manage cardiac arrest.”²⁰ The press release includes quotations from EMS personnel, demonstrating their belief that this device improves outcome from cardiac arrest, and also provides summary information from one of the studies reported in this issue.¹³ The press release also notes that “more than 235 EMS agencies and hospitals worldwide employ the AutoPulse.” Second quarter financial results for Zoll Medical Corporation released April 28, 2006, demonstrate an 8% increase in revenues and an 11% increase in sales in North America compared with the same quarter of the previous year.²¹ Furthermore, deliveries of the AutoPulse device increased 50% during the prior year (representing $2.4 million this quarter vs $1.6 million in the same quarter last year).²¹ It now seems this enthusiasm is premature, given that the effectiveness of the device likely depends on still-to-be-defined factors independent of the mechanical capabilities of the device. This illustrates the uneasy balance between forces promoting early adoption, including EMS personnel and perhaps physician enthusiasm unrelated to outcomes-based demonstration of efficacy as well as marketing efforts and business strategies, and the principle that new therapies should not be widely adopted, especially for diseases with substantial morbidity and mortality, until high-quality evidence accumulates that the new therapy is better than traditional treatments.

Can the differing conclusions of these 2 studies be reconciled? Although it is not possible to reach definitive conclusions until additional data are available, the best current information suggests that the degree of benefit or harm associated with use of the LDB-CPR device is influenced by the details of its use—perhaps including selection of the patient population with respect to presenting rhythm, time from cardiac arrest to initiation of CPR, and almost certainly time-to-deployment and the influence of deployment on time-to-defibrillation, when appropriate. To shed light on these questions, future comparative studies will need to pay particular attention to the definition and consistency of the method of use of the device, to measuring the multiple important time intervals with precision, and to ensuring the quality of the manual CPR administered in both trial groups.

Conducting high-quality clinical trials in patients with out-of-hospital cardiac arrest is extremely difficult because of the complexity of EMS systems, strong preconceived notions regarding effective therapies, and the almost ubiquitous presence of unmeasured confounding factors that profoundly influence outcomes for individual patients and treatment effects. However, only by aggressively identifying and rigorously testing promising new therapies will the outcomes improve for patients who experience sudden cardiac death.