Intravenous Drug Administration During Out-of-Hospital Cardiac Arrest:  A Randomized Trial FREE

Intravenous access and drug administration are integral parts of cardiopulmonary resuscitation (CPR) guidelines.¹ Millions of patients have received epinephrine during advanced cardiac life support (ACLS) with little or no evidence of improved survival to hospital discharge.^1- 2 The use of epinephrine is based on preclinical evidence of increased cerebral and coronary perfusion by redirected peripheral blood flow.^1- 2 Beneficial short-term effects of epinephrine have been shown in animal studies,^3- 5 but there is increasing concern for increased myocardial dysfunction^6- 7 and disturbed cerebral microcirculation after cardiac arrest.⁸ Epinephrine was an independent predictor of poor outcome in a large retrospective registry study,⁹ but this observational, nonrandomized study cannot prove a causal relationship. Despite its near-universal use, epinephrine has, to our knowledge, not been tested in a randomized controlled study with a no-drug comparison group.

If a negative association between epinephrine and survival is causal, it may be due to the drug or to inadequate CPR quality associated with drug administration. Drug administration includes time-consuming factors like establishing intravenous access, preparation, and administration of drugs and saline, thereby potentially removing focus from good-quality CPR. There are recent reports of poor-quality CPR and protocol adherence among professional CPR providers,^10- 11 and some consider intubation and intravenous access more important than giving good-quality chest compressions.¹² With inadequate CPR quality, effects of drugs administered peripherally also may be diminished or absent.¹³ Because there are no randomized controlled studies showing improved survival to hospital discharge with any drugs routinely administered during CPR, we concluded such a study was warranted.

In this prospective, randomized controlled trial of intravenous drug administration during out-of-hospital cardiac arrest, we compared outcomes for patients receiving standard ACLS with intravenous drug administration (control) and patients receiving ACLS without intravenous drug administration (intervention).

The city of Oslo has a single-tiered emergency medical service system administered by the Oslo University Hospital for a population of 540 000. On weekdays between 7:30 AM and 10:00 PM, an ambulance staffed by 2 paramedics and an anesthesiologist functions on the same level as the regular paramedic-staffed ambulances. Until January 2006, ACLS was performed according to the International Guidelines 2000,¹⁴ with the modification that patients with ventricular fibrillation received 3 minutes of CPR before the first shock and between unsuccessful series of shocks.¹⁵ The European Resuscitation Council Guidelines for Resuscitation 2005¹⁶ were implemented in January 2006, incorporating this same modification of 3-minute periods of CPR. Defibrillators in manual mode are used and endotracheal intubation is standard for securing the airways. Two ambulances are routinely dispatched for suspected cardiac arrest. The physician-staffed ambulance is dispatched whenever available.

All hospitals in Oslo have goal-directed postresuscitation protocols including therapeutic hypothermia regardless of initial rhythm or arrest etiology.¹⁷ A prehospital 12-lead electrocardiogram is routinely transmitted to the cardiologist on call after return of spontaneous circulation (ROSC). If coronary angiography is indicated for possible percutaneous coronary intervention, patients are transported directly from the scene to 1 of 2 university hospitals (Oslo University Hospital, Ullevaal and Rikshopitalet) with this capacity 24 hours per day.

All patients older than 18 years with nontraumatic, out-of-hospital cardiac arrests between May 1, 2003, and April 28, 2008, were randomized by ambulance personnel on-site. Simple randomization occurred directly after ambulance personnel confirmed the cardiac arrest and then opened the sealed envelopes provided by the investigators. Patients were randomized to receive either ACLS with access to intravenous drug administration (intravenous group) or ACLS without access to intravenous drug administration (no intravenous group). In the no intravenous group, intravenous access was to be established 5 minutes after ROSC, and drugs could then be given if indicated.

Exclusion criteria were (1) cardiac arrest witnessed by ambulance crew because these patients almost always have an intravenous needle in place at the time of the cardiac arrest, (2) resuscitation initiated or interrupted by physicians outside of the ambulance team, or (3) cardiac arrest induced by asthma or anaphylactic shock (which were the last criteria added in October 2006). The study was approved by the regional ethics committee. Informed consent for inclusion was waived as decided by this committee, but was required from survivors with 1-year follow-up.

Standard defibrillators (LIFEPAK 12 Physio-Control, Medtronic, Redmond, Washington) were used. Electrocardiograms with transthoracic impedance signals from these defibrillators were routinely transferred to a server at the National Competence Center for Emergency Medicine (Oslo, Norway) following cardiac arrest. Utstein cardiac arrest forms¹⁸ routinely completed by paramedics were submitted to the study supervisor along with a copy of the ambulance run sheet. Automated, computer-based dispatch center time records supplemented ambulance run sheets with regard to response intervals. For admitted patients, additional hospital records were obtained.

All trial data were documented according to the Utstein style.¹⁸ The primary end point was survival to hospital discharge. Secondary outcomes were 1-year survival, survival with favorable neurological outcome (using cerebral performance categories from 1 to 4),¹⁸ hospital admission with ROSC, and quality of CPR (ie, chest compression rate, pauses, and ventilation rate). The study was monitored annually with interim analysis by an external researcher who did not reveal any results to the investigators.

Data from each case were viewed and annotated using CODE-STAT 7.0 (Physio-Control, Medtronic) for detection of ventilations and chest compressions by changes in transthoracic impedance. Written information from patient report forms and locally adapted Utstein style forms were compared with typical changes in CPR patterns as shown using CODE-STAT 7.0. Initial rhythm assessment registered on patient report forms were confirmed by these electrocardiographic recordings if possible. Time without spontaneous circulation, time without compressions during time without spontaneous circulation (hands-off time), preshock pauses, compression rate and actual number of compressions, and ventilations per minute were calculated for each episode. Hands-off ratio is defined as hands-off time divided by total time without ROSC. Electrocardiographic analysis was performed by 1 researcher (T.M.O.).

Initial power analysis was based on survival statistics for the Oslo emergency medical service system and assumed that the survival rate would be doubled among patients not receiving epinephrine, as described previously in an observational study.⁹ With a projected survival rate of 7% in the intravenous group and 14% in the no intravenous group, 900 patients provided a power level of 91.4% with a type I error of 5%.¹⁹

Analysis was performed on an intention-to-treat basis regardless of which treatment was actually given. Patients who were initially randomized, but were later found to meet predefined exclusion criteria were not included in the intention-to-treat analysis. Demographic and clinical data are presented as means with 95% confidence intervals (CIs), medians with ranges, or proportions. Crude effects between the 2 trial groups and survival were quantified by odds ratios (ORs) with 95% CIs. The χ² test for contingency tables with different degrees of freedom was used to detect associations between categorical independent variables. For continuous variables, the t test was used for normally distributed data and the Mann-Whitney test was used for nonnormally distributed data.

Confounders were identified and quantified by using the Mantel-Haenszel test for both short-term and long-term survival, and subsequent manual backward-elimination procedures were performed. Correlations between potential confounders were investigated. Comparison of Kaplan-Meier survival curves was obtained using the Breslow and log-rank test statistics for short-term and long-term survival, respectively.^20- 21

Two-sided P values of less than .05 were considered significant. The statistical analyses were performed using the software packages SPSS version 15.0 and SamplePower version 2.0 (SPSS Inc, Chicago, Illinois) and Egret version 2.0.31 (Cytel Software Corporation, Cambridge, Massachusetts).

Resuscitation was attempted in 1183 patients who experienced cardiac arrest during the study period, and 851 of 946 eligible patients were successfully randomized with 418 patients in the intravenous group and 433 patients in the no intravenous access group. For reasons listed in Figure 1, 95 eligible patients were not randomized and further randomization and inclusion details are illustrated. Eligible, nonrandomized patients did not differ significantly from randomized patients with regard to demographic characteristics and outcomes.

Baseline demographic characteristics and CPR-quality parameters are listed in Table 1. Defibrillation was attempted in more patients in the intravenous group compared with the no intravenous group (47% vs 37%, respectively; OR, 1.16 [95% CI, 0.74-1.82]). More defibrillation shocks were delivered to those who received defibrillation in the intravenous group compared with the no intravenous group (median, 3 [range, 1-22] vs 2 [range, 1-26], respectively; P = .008). Both groups had adequate and similar CPR quality with few chest compression pauses (median hands-off ratio, 0.15 for the intravenous group and 0.14 for the no intravenous group) and the compression and ventilation rates were within the guideline recommendations (Table 1).

In the intravenous group, 44 of 418 patients (10.5%) survived to hospital discharge vs 40 of 433 (9.2%) in the no intravenous group (OR, 1.16; 95% CI, 0.74-1.82; P = .61). Survival with favorable neurological outcome was 9.8% for the intravenous group and 8.1% for the no intravenous group (OR, 1.24; 95% CI, 0.77-1.98; P = .45). Short-term survival was significantly better in the intravenous group than in the no intravenous group with 40% vs 25%, respectively, achieving ROSC (OR, 1.99; 95% CI, 1.48-2.67; P < .001), 43% vs 29% admitted to the hospital (OR, 1.81; 95% CI, 1.36-2.40; P < .001), and 30% vs 20% admitted to the intensive care unit (ICU) (OR, 1.67; 95% CI, 1.22-2.29; P = .002) (Table 2). In-hospital treatments, including therapeutic hypothermia and percutaneous coronary intervention, were equally distributed between the 2 groups. There were no differences in cause of death among patients admitted to the ICU and most deaths were due to brain damage (Table 2).

Patients were divided into 2 predefined subgroups based on their initial rhythms (Table 3). In patients with an initial rhythm of ventricular fibrillation or pulseless ventricular tachycardia, there were no differences in short-term or long-term outcomes. In the subgroup with nonshockable rhythms (initial rhythm of asystole or pulseless electrical activity), the ROSC rate was 3-fold higher with intravenous treatment (P < .001), but there was no difference in long-term outcome because the survival rate among those admitted to the ICU tended to be almost 3 times higher in the no intravenous group (P = .07; Table 3).

A public cardiac arrest location, response interval, and initial ventricular fibrillation were identified as potential confounders and were included in the logistic regression analysis. Multivariate logistic regression analyses for short-term survival (admitted to the ICU) and long-term survival (discharged from the hospital) were performed. After adjustment for confounders, patients in the intravenous group had a nonsignificant 15% increased chance of surviving to hospital discharge (adjusted OR [AOR], 1.15; 95% CI, 0.69-1.91) compared with patients in the no intravenous group. Patients with ventricular fibrillation or pulseless ventricular tachycardia as the initial rhythm had a 10-fold improvement in long-term survival (AOR, 10.47; 95% CI, 5.47-20.03). Patients with bystander-witnessed cardiac arrests or cardiac arrests in public places had a 2-fold improvement in long-term survival (AOR, 2.13 [95% CI, 1.02-4.45] and AOR, 2.03 [95% CI, 1.19-3.44], respectively), whereas the odds of long-term survival decreased by 17% for each minute of prolonged response interval (AOR, 0.83; 95% CI, 0.77-0.90). When adjusted for the same confounding factors, survival to ICU admission was higher for patients in the intravenous group (AOR, 1.78; 95% CI, 1.26-2.51).

The cumulative postcardiac arrest survival rate at 7 days was 14.6% (95% CI, 11.3%-17.9%) for patients in the intravenous group vs 12.8% (95% CI, 9.7%-15.9%) for patients in the no intravenous group, 11.3% (95% CI, 8.4%-14.2%) vs 8.8% (95% CI, 6.1%-11.5%), respectively, at 1 month, and 9.8% (95% CI, 6.9%-12.7%) vs 8.4% (95% CI, 5.9%-10.9%) at 1 year (Figure 2). Short-term survival was significantly higher for patients in the intravenous group compared with patients in the no intravenous group (Breslow P = .004), although there was no difference in long-term survival (log-rank P = .23)

Our results represent the first attempt, to our knowledge, to evaluate the effect of intravenous access and intravenous drug administration on outcome in patients with an out-of-hospital cardiac arrest. Short-term survival was higher in the intravenous group, but these nearly universally applied interventions were not associated with a statistically significant improvement in survival to hospital discharge.

Administration of intravenous drugs did not appear to interfere with the quality of CPR. Ambulance personnel delivered good-quality CPR with few pauses and with rates within guideline recommendations¹ in both groups. This is important because potential improvements in intravenous medication administration during ACLS will not need to overcome an intrinsic tendency to degrade CPR.

We did not confirm the previous observational finding that intravenous epinephrine was an independent predictor for poor outcome.⁹ Our results are consistent with a multicenter study by Stiell et al²² that found no difference in survival after implementing intravenous drug administration during out-of-hospital cardiac arrest (OR, 1.1; 95% CI, 0.8-1.5).

Without differences in the predefined primary outcome, patients in the intravenous group received more defibrillations, were resuscitated for a longer period, and more frequently had ROSC. With similar and adequate CPR quality, this is likely due to the pharmacological effects of the drugs used (epinephrine, atropine, and/or amiodarone). This finding is consistent with previous animal studies with epinephrine,^6- 7 and clinical studies evaluating the effects of amiodarone,²³ atropine,²⁴ and even high-dose epinephrine,²⁵ all of which documented improved short-term effects without improving long-term outcomes. While epinephrine can produce more spontaneously beating hearts in animal models, it is also associated with increased postresuscitation myocardial dysfunction that might partly explain these clinical observations.^6- 7 Negative postresuscitation effects of epinephrine also are reported to be more prominent after longer, more clinically relevant cardiac arrest periods (eg, 4-6 minutes) than short cardiac arrest periods (eg, 2 minutes).⁷ Moreover, an experimental study has recently documented detrimental effects of epinephrine on cerebral microcirculation.⁸

The clinical implications of an increased ROSC rate in the intravenous group are difficult to interpret. Should improved short-term outcome be regarded as unfulfilled potential that might be addressed with better post-ROSC care, or unproductive resuscitation of patients whose vital organ injury makes them unlikely candidates for long-term survival? In the present study, most patients who died in the hospital after initial successful resuscitation in both groups had severe cerebral damage. If present pharmacological interventions only facilitate cardiac resuscitation in patients who will ultimately experience irreversible cerebral damage, this may cause an additional burden on already overburdened ICUs.

However, long-term survival cannot be achieved without first restoring circulation. Improved brain-directed postresuscitation treatment might at some point prevent irreversible cerebral damage and increase survival. At present, the only established brain-directed treatment is therapeutic hypothermia,^26- 27 and the rate of which was high in both groups (71% and 72%). It is possible that for some patients in our study with early postresuscitation cardiac death, advanced options such as mechanical chest compression devices,²⁸ extracorporeal membrane oxygenation,²⁹ or left ventricular assist devices³⁰ could enable corrective treatment of underlying causes and theoretically improve survival.

The results of our study highlight the question of whether patients presenting with initial shockable rhythms and nonshockable rhythms should be treated differently. Initial shockable rhythm was a potential effect modifier in our statistical analysis, indicating that the degree of benefit or harm of intravenous drug administration during cardiac arrest may depend on the presenting rhythm. No differences in outcome were found for patients with shockable rhythms, while patients with nonshockable rhythms had higher rates of ROSC in the intravenous group, but an opposite tendency toward a lower rate of survival to hospital discharge among those admitted to the hospital. This suggests that late toxicity after intravenous drug administration contributes importantly to the poor outcomes of these patients.

Several studies have identified dissimilar etiologies in subgroups with shockable and nonshockable rhythms,^31- 33 and it seems reasonable that differences in treatment strategies will emerge.³⁴ Retrospective subgroup analysis for cardiac arrest times (<5 minutes, 5-10 minutes, or >10 minutes) did not reveal any suggestive information either alone or combined with initial rhythm (data not presented but available from authors upon request). However, our study was not powered for formal subgroup analysis and no conclusions should be drawn.

The present data indicating good-quality CPR in both groups suggest that the lack of improved long-term outcome with ACLS with intravenous drug administration cannot be explained by poor-quality CPR.¹³ This does not exclude the possibility that other drug regimens might improve outcome. Early administration, as recently advocated,^35- 36 must be evaluated in systems with shorter ambulance response intervals or other intravenous drug regimens and priorities that are different from the present guidelines.

Our study has several limitations. First, ambulance personnel could not be blinded to the randomization. Closely related to this, only patients who were randomized to the no intravenous group could be monitored with regard to protocol compliance. If intravenous drugs were administered to a patient in the no intravenous group, violation of the study protocol could be documented. If intravenous drugs were not administered to a patient in the intravenous group, several valid reasons could exist, such as rapid ROSC. We have no reason to believe that personnel refrained from establishing intravenous access under the pretense that the procedure was unsuccessful. The ambulance personnel involved were strongly committed to testing the hypothesis presented, but we cannot totally rule out possible bias toward procedures such as intravenous access and administration of drugs, which have been important in Norwegian culture for decades.

Second, quality of CPR could only be assessed in 75% of cases. Still, this is, to our knowledge, the first clinical intervention study reporting CPR quality data, and no significant differences were found between these data and those unavailable for analysis. Also, we do not have reliable time points for drug administration. Paramedics in the Oslo emergency medical service system are highly trained and both the guidelines and training emphasize early intravenous access and drug administration and intubation with the shortest possible pauses in chest compressions.

Third, this is a single center study and the results may not be generalized to systems with different training, infrastructure, treatment protocols, or quality of CPR. Fourth, while time from cardiac arrest to the initiation of ACLS is important for patient survival, the estimated time of cardiac arrest is imprecise and one-third of the cardiac arrests were unwitnessed. This variable is therefore not included in the analysis. Only the emergency medical service response interval was included.

Finally, a type II error cannot be ruled out. Although based on the best available evidence at the time,⁹ the power analysis was, in retrospect, optimistic in assuming a doubling in survival for the patients in the no intravenous group. For the observed difference between the groups to be statistically significant, a sample size of 14 000 patients would be needed. Because this sample size has not been considered inappropriate in cardiovascular interventions, our results could be background for such a large study that could be positive for intravenous access and drug administration. At a minimum, our results indicate that clinical equipoise exists on the efficacy of intravenous drugs in the treatment of cardiac arrest and that more definitive trials could be ethically undertaken. Alternatively, the poor survival rates after cardiac arrest, which do not seem to be significantly improved by intravenous drug administration, indicate that research should be directed at new pharmacological interventions that hold promise of greater effect.

Despite improved short-term survival among patients randomized to receive intravenous access and drug administration, these nearly universal interventions were not associated with a statistically significant improvement in survival to hospital discharge. Larger trials examining resuscitation without intravenous access and drug administration, as well as of existing or new drugs, appear to be justified.