Pilot Randomized Clinical Trial of Prehospital Induction of Mild Hypothermia in Out-of-Hospital Cardiac Arrest Patients With a Rapid Infusion of 4°C Normal Saline
Background— Although delayed hospital cooling has been demonstrated to improve outcome after cardiac arrest, in-field cooling started immediately after the return of spontaneous circulation may be more beneficial. The aims of the present pilot study were to assess the feasibility, safety, and effectiveness of in-field cooling.
Methods and Results— We determined the effect on esophageal temperature, before hospital arrival, of infusing up to 2 L of 4°C normal saline as soon as possible after resuscitation from out-of-hospital cardiac arrest. A total of 125 such patients were randomized to receive standard care with or without intravenous cooling. Of the 63 patients randomized to cooling, 49 (78%) received an infusion of 500 to 2000 mL of 4°C normal saline before hospital arrival. These 63 patients experienced a mean temperature decrease of 1.24±1°C with a hospital arrival temperature of 34.7°C, whereas the 62 patients not randomized to cooling experienced a mean temperature increase of 0.10±0.94°C (P<0.0001) with a hospital arrival temperature of 35.7°C. In-field cooling was not associated with adverse consequences in terms of blood pressure, heart rate, arterial oxygenation, evidence for pulmonary edema on initial chest x-ray, or rearrest. Secondary end points of awakening and discharged alive from hospital trended toward improvement in ventricular fibrillation patients randomized to in-field cooling.
Conclusions— These pilot data suggest that infusion of up to 2 L of 4°C normal saline in the field is feasible, safe, and effective in lowering temperature. We propose that the effect of this cooling method on neurological outcome after cardiac arrest be studied in larger numbers of patients, especially those whose initial rhythm is ventricular fibrillation.
Received August 1, 2006; accepted April 9, 2007.
After resuscitation from cardiac arrest, brain injury is a major source of morbidity and mortality. Most patients who are resuscitated from cardiac arrest never awaken.1–4 Despite delays of 4 to 8 hours in its initiation, mild hypothermia (32°C to 34°C) induced in patients resuscitated from out-of-hospital ventricular fibrillation (VF) has been shown to improve neurological recovery and survival.5,6
Clinical Perspective p 3070
Results from animal models suggest that the effectiveness of mild hypothermia could be improved if initiated as soon as possible after return of spontaneous circulation (ROSC).7–9 Bernard et al10,11 have hypothesized that early initiation of rapid cooling, preferably in the field soon after ROSC, will have the maximum benefit in both neurological outcome and survival.
Rapid infusion of cold intravenous fluids is an attractive option because it would be easy to initiate in the field after ROSC. In 2 small hospital-based pilot studies of resuscitated cardiac arrest patients, temperature decreased by 1.7°C in 22 patients treated with cold lactated Ringer’s solution11 and by 1.7°C in 17 patients infused with cold normal saline12 without clinically significant hemodynamic changes or electrolyte abnormalities. In 1 trial, echocardiographic studies demonstrated no significant effect on left ventricular function, left atrial filling pressures, pulmonary artery pressure, and central venous pressures.12 These trials, however, were conducted after hospital admission, and safety and effectiveness of this cooling method initiated in the field have not been established.
In the present pilot study, we randomized 125 patients to receive standard care with or without the induction of mild hypothermia using a rapid infusion of 2 L of 4°C normal saline as soon as paramedics had resuscitated the patient from out-of-hospital cardiac arrest. The first aim of the present study was to examine the feasibility of cooling resuscitated patient in the field before their arrival at the emergency department. To assess safety, the second aim, we examined whether field cooling was associated with adverse effects on rearrest, on field or hospital arrival hemodynamics, on oxygenation or pulmonary edema, and on hospital variables such as prolonged length of stay in the hospital or increased mortality. Finally, in our third aim, temperature changes were used as a measure of efficacy.
This pilot trial was conducted in Seattle, Wash, and involved Medic One with its 78 paramedics rotating among 7 paramedic units that serve the city and its 9 acute care hospitals. Patients were eligible if they were ≥18 years of age, were resuscitated by paramedics from nontraumatic out-of-hospital cardiac arrest, had an esophageal temperature of ≥34°C, were intubated, had intravenous access, and were unresponsive. Cardiac arrest was defined as being unconscious as a result of a sudden pulseless collapse; ROSC was a return of a palpable pulse in a patient with cardiac arrest. The inclusion and exclusion criteria are detailed in Figure 1. Except for trauma, all causes of cardiac arrest were considered, including those with initial rhythms of VF or asystole or the state of pulseless electrical activity. Patients meeting all of the eligibility criteria listed in Figure 1 were randomized to receive standard care alone or standard care plus the induction of mild hypothermia. Paramedics called the emergency department physician at Harborview Medical Center to verify eligibility and to learn treatment assignment. The emergency room physician opened sequentially numbered envelopes that randomized patients to either receive or not receive cooling. Randomization was carried out in balanced blocks of 4.
The clinical trial required equipment to be added to the paramedic units and some changes to be made in the resuscitation protocols. Each of the paramedic units was equipped with refrigerators capable of storing several 1-L bags of normal saline at 4°C. In addition, in the months before the start of the present study, paramedics began to place esophageal temperature probes (Acoustascope Esophageal Stethoscope with temperature sensor, Level One, Rockland, Me) after tracheal intubation in all resuscitated out-of-hospital cardiac arrest patients. As part of the present study, paramedics were to record temperatures at randomization and at hospital arrival using a portable temperature recorder (YSI Precision 4000 A Thermometer, YSI Corp, Dayton, Ohio).
For resuscitated eligible patients randomized to cooling, paramedics administered intravenously up to 2 L of 4°C normal saline, pancuronium (7 to 10 mg), and diazepam (1 to 2 mg), as in the prior pilot study of patients treated in hospital.12 Seattle Medic One paramedics already use intravenous pancuronium and diazepam in the field but not for this indication. The saline was infused through a peripheral intravenous line, 18-gauge or larger, using a pressure bag inflated to 300 mm Hg. We did not adjust the amount of 4°C normal saline to body weight. If the patient suffered another cardiac arrest during transport, standard resuscitation protocols were started, and saline infusion was stopped. Patients randomized to standard care with or without cooling were otherwise treated the same according to Medic One resuscitation protocols.
Paramedics transported patients to acute care hospitals in Seattle and provided information sheets describing the study and encouraging continued cooling in the hospital. Once at the hospital, patients were no longer treated as part of the study protocol but rather according to their physicians’ preferences. For example, emergency department staff at some hospitals stopped the 4°C normal saline infusions that had not been completed during transport. In addition, at some hospitals, cooling was initiated or continued regardless of what had been done in the field. For example, during the course of the present study, patients cared for at 1 facility after cardiac arrest were routinely cooled for 24 hours. That institution receives approximately half of transported cardiac arrest patients. During the course of the study, 2 other hospitals also began to cool patients admitted after cardiac arrest; the remaining hospitals did not.
From standard run reports that paramedics complete, we collected data on the prehospital resuscitation: initial blood pressure, heart rate, use of pressors, rearrest, or recurrent VF. From hospital records, we collected data on demographics; whether cooling was initiated or continued in the hospital; blood pressure, heart rate, and pulse oximetry data during first 12 hours; first arterial blood gas; first anterior-posterior chest x-ray interpretations (we abstracted data on whether the attending radiologist mentioned pulmonary edema, pulmonary congestion, hilar abnormalities, cardiomegaly, or pleural effusion); use of intravenous diuretics; and use of pressors (dobutamine, dopamine, norepinephrine, epinephrine, or phenylephrine). We determined the number of days to discharge or death and whether awakening occurred, which was defined as the patient following commands or having comprehensible speech. Times were calculated in days, and we used both the date and time of arrest, as well as the date and time of death and time of discharge.
All analyses were conducted on the basis of intention to treat. Using SPSS version 13.0 (SPSS Inc, Chicago, Ill), we analyzed differences between the 2 groups with Student t test for continuous variables and the χ2 statistic for categorical variables. In situations in which continuous variables were not normally distributed, the Wilcoxon rank sum test was used to determine statistical significance. Two-tailed tests were performed with α=0.05. Continuous values were presented as mean±SD or as medians and ranges. ANOVA was used to examine the association between temperature change and fluid volume. We planned subgroup analyses by initial rhythm: VF versus other. To further assess the effect of hospital cooling on the association between field cooling and survival status at hospital discharge, logistic regression was performed using these exploratory variables: field cooling, yes or no; hospital cooling, yes or no; and an interaction term (product of the field and hospital cooling variables, with 1=field and hospital and 0=all others).
Because we hypothesized that, for the treatments to be effective, the infusions had to be given as soon as possible after circulation was restored, informed consent could not be obtained before a patient entered into the study and was thus waived. Human subject committees at the University of Washington and all the acute care hospitals in Seattle reviewed and approved the study. As required by federal regulations concerning the waiver of consent, we obtained an Investigational New Drug number from the Food and Drug Administration. We also informed all of the cardiologists, directors of emergency departments, and directors of intensive care units of acute care hospitals in Seattle about the study. Whenever possible, study personnel contacted the patient’s family to explain the study and seek written informed consent for continued participation in the study. Because no study interventions occurred after admission, consent was sought only to allow study personnel to review information collected by Seattle Medic One and the receiving hospitals. If a randomized patient died before this contact, the patient’s family was still informed about the study.
The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.
The study began on November 15, 2004, and the 125th patient was enrolled on February 5, 2006. Over this interval, paramedics attended 559 cardiac arrests (Figure 1). Most patients were not eligible because cardiopulmonary resuscitation was not successful. Reasons for not being eligible or enrolled are listed in Figure 1. Sixty-five patients who were eligible to be randomized, however, were not enrolled. Paramedics failed to consider enrollment for 23 patients. Twenty-four patients were classified in the field as being too unstable for enrollment. Seventeen of these 24 patients lost pulses within 5 minutes of ROSC and could not be randomized. They all subsequently died in the field or in the emergency department. Baseline characteristics are shown in Table 1 and were not significantly different between the 2 treatment groups. Patients were admitted to 7 of the 9 acute care hospitals in Seattle.
None of the 62 patients randomized to standard care alone had cooling induced in the field. On the other hand, not all of the 63 patients randomized to cooling received 4°C normal saline in the field. Eight patients did not receive any fluid, 6 patients received <500 mL, 37 patients received between 500 mL and 2 L, and 12 patients received the full 2 L. A greater decrease in temperature was associated with a greater amount of cold fluid infusion (Figure 2). The reasons the full 2 L was not administered included recurrent arrest, death in the field, and lack of time before hospital arrival to complete the 2-L infusion. In most of these patients, the infusion was not completed by the time of hospital arrival, and the infusion was stopped at the discretion of emergency department personnel.
The primary efficacy outcome was the temperature difference: at hospital arrival minus at randomization in the field. In the patients not randomized to cooling, paramedics failed to record temperatures at randomization in 8 patients and at hospital arrival in 18 (3 died in the field), so temperature differences were available in only 36 patients who received standard care only. For those randomized to cooling, paramedics failed to record temperatures at randomization in 1 patient and at hospital arrival in 8 (7 died in the field), so temperature differences were available in 54 patients. Although the temperatures at randomization did not differ between the treatment groups, those at admission did differ significantly, as did the temperature differences between randomization and hospital arrival (Table 2).
Prehospital and hospital deaths in both treatment groups were similar (Figure 1 and Table 3). Field cooling was not associated with an increase in the time the patient spent in the field before hospital arrival (Table 4). Prehospital (Table 4) and early hospital (Table 5) variables were not significantly different by field cooling status except for the early hospital variables of higher heart rate, higher blood pressure, and acidemia being more common and cardiomegaly on chest x-ray being less common in those randomized to field cooling compared with those not. The changes in systolic blood pressure and heart rate from the times of randomization to arrival at the emergency department were similar for both control and cooled patients (Table 5).
For hospitalized patients, days to awakening, days to discharge, and days to death were not significantly different between the 2 groups (Table 5). The range from arrest to death was 0.2 to 89 days, and the range from arrest to discharge was 2.9 to 32 days. Almost 70% of deaths occurred within 2 days of cardiac arrest, and >40% of discharges occurred within 10 days of arrest.
There was a trend toward improved survival to discharge in the patients randomized to field cooling when the initial rhythm was VF (Table 3). The reverse was evident when the initial rhythm was not VF. However, none of these differences was statistically significant. Of all of the discharged patients, 2 patients in each treatment group had severe neurological deficits. Only 3 died in the hospital after awakening after cardiac arrest. Considering the outcome of ever awakening in patients resuscitated from out-of-hospital cardiac arrest with the initial rhythm of VF, 20 of 29 patients (69%) randomized to field cooling awakened compared with 10 of 22 VF patients (45%) not randomized to cooling (P=0.15, 2-sided Fisher exact test). On the other hand, in those patients with initial rhythms not VF, only 3 of 34 patients (9%) randomized to field cooling awakened compared with 9 of 40 patients (23%) not randomized to cooling (P=0.13, 2-sided Fisher exact test).
Of the 97 patients who were admitted to hospital, 60 were treated, at the discretion of their treating physician, with hypothermia induced through the use of surface cooling. Given that hospital cooling could either confound or modify the effect of field cooling on survival to hospital discharge, we conducted exploratory analyses on the 97 admitted patients using logistic regression with the outcome variable being survival to hospital discharge. Overall, the unadjusted odds ratio for field cooling was 1.25 (95% CI, 0.55 to 2.82), an insignificant trend toward improved survival to hospital discharge. In a multiple logistic regression model, we used the exploratory variables of field cooling, hospital cooling, and an interaction variable. In patients who received field cooling alone, the odds ratio for survival to hospital discharge was 1.92 (95% CI, 0.46 to 8.0) adjusted for hospital cooling and the interaction term; for those receiving hospital cooling alone, the odds ratio was 0.91 (95% CI, 0.28 to 2.96) adjusted for field cooling and the interaction term. We did not detect a significant interaction between the field cooling and hospital cooling variables. When we adjusted for the effects of hospital cooling, the odds ratio for survival to hospital discharge for the field cooling group increased slightly from 1.25 to 1.38 (95% CI, 0.58 to 3.29).
In this prehospital pilot randomized clinical trial, we evaluated whether the strategy of using a rapid infusion of up to 2 L of 4°C normal saline in the field after resuscitation from out-of-hospital cardiac arrest would be effective in lowering the temperature before arrival at the hospital. We demonstrated the feasibility of this approach and found it to be effective in inducing mild hypothermia by hospital arrival. These pilot data suggest that field cooling is not associated with adverse effects on blood pressure, heart rate, or pulmonary edema.
On the basis of our previous in-hospital study and the experience from Bernard et al11 and Kim et al,12 a temperature decrease of 1.7°C to 2.0°C should be expected with an infusion of 2 L cold fluid. We first asked whether this approach would be feasible after resuscitation in the field. Seattle paramedics were able to infuse between 500 and 2000 mL fluid in 78% of the patients randomized to receive fluid before hospital arrival. As expected, more fluid was associated with a greater temperature decrease. The main limitation in the amount of fluid infusion was a result of the short transport time and the discontinuation of fluid infusion by emergency department personnel. Our field intervention was not designed to be continued in the hospital. If the paramedics had infused the full 2 L in all patients randomized to cooling, the temperature decrease may well have approached 1.6°C to 1.7°C, as expected from prior studies, instead of the 1.2°C observed.
We had several missing temperature values at hospital arrival in those not randomized to field cooling. This failure to record arrival temperature in a number of control patients was most likely due to the paramedics simply overlooking that step in some patients who were not actually cooled. Having this information would be unlikely to change the results because control patients would have to have had spontaneously and markedly decreased temperatures to negate the temperature differences observed between the 2 groups. Missing values could have been averted if continuous portable temperature recording were available, eliminating the need for the paramedics to record temperatures. Not all emergency departments had the capability to measure temperatures with an esophageal temperature probe and instead used a tympanic thermometer. Our previous experience with measuring both the esophageal and tympanic values indicated that these temperature values often did not agree; therefore, we did not use hospital temperatures (often measured by tympanic thermometer) data to replace missing emergency department arrival temperature. Continuous and standard temperature recording in field and in hospital will become important if field cooling becomes more widespread.
A major concern in using a rapid infusion of cold fluid in the field is the possibility of inducing pulmonary edema in resuscitated patients. These pilot data suggest that the rate of pulmonary edema was not significantly different between the 2 groups, confirming results of prior studies based on fluid administration after hospital arrival. Patients receiving field cooling had a higher heart rate and systolic blood pressure at hospital arrival, which does not appear to be clinically significant. Because the heart rate at randomization tended to higher in the cooling group, we also examined pair differences between randomization and hospital arrival heart rate and systolic blood pressure. These paired differences are not significant. In addition, the incidence of cardiomegaly as seen on chest x-ray was actually less in patients randomized to field cooling, suggesting that field cooling did not worsen cardiac function.
We observed in the 2 treatment groups a significant difference in the first measured pH arterial blood gas, suggesting that field cooling was associated with a greater degree of acidemia without changes in arterial Pco2. One possible explanation might be related to an increase in chloride load during the infusion of cold normal saline. We note, however, that such a shift in pH has not been seen in previous reports with infusion of cold fluid11,12 and should be reexamined in subsequent studies. Although these results suggest that this method of cooling is safe, confirmation is needed in a larger, more adequately powered study.
In the Bernard et al6 study of mild hypothermia in resuscitated VF patients, serum potassium levels decreased with the induction of surface cooling and increased during the rewarming stage. Mild hypothermia also has been shown to possibly increase serum glucose levels as a result of reduced insulin secretion from the pancreas. In the present study, we did not systematically collect data on ECGs or potassium and glucose levels, and we cannot comment on changes that might have been related to the intervention. However, in our in-hospital study of mild hypothermia and in other studies, no adverse effects of cooling on serum glucose or potassium levels were detected.12,13
Sixty-five apparently eligible patients (34%) were not enrolled. Recurrent cardiac arrest during transport to the hospital developed in 17 of these patients (26%), a rate quite similar to that for the cooled (24%) or control (21%) patients.
One interesting aspect of our results suggests that the infusion of up to 2 L fluid may not have similar benefits when patients are stratified by initial cardiac rhythm. In patients with VF, field cooling had a trend toward improved outcome, even though this intervention was started in the field and despite the fact that not all of these patients received the full 2-L infusion. Thus, field cooling was associated with a higher proportion of VF patients discharged alive compared with those who received standard care only (66% versus 45%). On the other hand, in patients who had an initial rhythm of pulseless electric activity or asystole, field cooling was associated with a lower proportion of patients discharged alive (6% versus 20%). These trends are preliminary and statistically not significant in this relatively small sample size. However, the findings do justify a larger study of this intervention, with reexamination of the influence of the presenting rhythm.
Despite recommendations by International Liaison Committee on Resuscitation and the American Heart Association, hospitals have been slow in instituting cooling of resuscitated VF patients.14 During the present trial, 60 of 97 admitted patients (62%) received hospital cooling regardless of field cooling. One positive outcome of the present pilot study has been that Seattle-area hospitals are now more aware of the use of mild hypothermia in resuscitated cardiac arrest patients. Our preliminary analyses did not suggest that the effect of field cooling on outcomes was either confounded or modified by hospital cooling, although these questions need to be addressed in larger studies.
We have focused on field hypothermia for at least 3 reasons. First, animal data suggest that hypothermia is more effective if induced as soon as possible after ROSC.7–9 Second, the initiation of hypothermia in the field is attractive in that paramedics respond to virtually all out-of-hospital cardiac arrests and that their application of hypothermia may be readily standardized. Third, a simple means to induce hypothermia in the field is available with the infusion of cold intravenous fluids.
The present study has a number of limitations. These include the absence of some follow-up temperatures, limited volume of infusion of cold normal saline, and differences in hospital policy for hospital cooling of admitted cardiac arrest patients. Given that the present trial was designed as a pilot and feasibility study, our sample size is not sufficient to render a conclusion on the effect of field cooling on survival or neurological outcome.
It was not possible to blind the Seattle paramedics to treatment. Study personnel during data collection and analysis could not be entirely unaware of treatment assignment; however, many of the outcome variables (eg, death, heart rate) were objectively measured and are less susceptible to bias. Ideally, the control group should have received an equal amount of noncooled intravenous fluid. Without this control, we cannot be sure that any differences in outcomes relate strictly to differences in temperatures. Similarly, the control group did not receive pancuronium and diazepam, although a previous trial showed no beneficial effects of diazepam on outcome after cardiac arrest.4
The infusion of 4°C normal saline in the field appears to be an effective method for inducing mild hypothermia in resuscitated out-of-hospital cardiac arrest patients. The administration of cold fluid was not associated with adverse effects and was associated with a significant temperature decrease by hospital arrival. A larger clinical study is warranted to determine whether field cooling is associated with improved neurological outcome and survival in resuscitated out-of-hospital cardiac arrest patients.
We thank Shirley Whitkanack for obtaining consent and for patient follow-up and Michele Prock for data entry. We also acknowledge the outstanding efforts of the Seattle Fire Department paramedics and the emergency room physicians at Harborview Medical Center who advised paramedics on treatment randomization. The Data Safety Monitoring Committee was made up of Margaret Neff, MD; Kyra Becker, MD; Tina Chang, MD; Nancy Temkin, PhD; and Earl Sodeman, Deputy Chief Seattle Fire Department.
Sources of Funding
This work was supported by a grant from the Medic One Foundation and National Institutes of Health (NIH) grant HL04346 (F.K.).
Longstreth WT, Jr., Fahrenbruch CE, Olsufka M, Walsh TR, Copass MK, Cobb LA. Randomized clinical trial of magnesium, diazepam, or both after out-of-hospital cardiac arrest. Neurology. 2002; 59: 506–514.
Abella BS, Zhao D, Alvarado J, Hamann K, Vanden Hoek TL, Becker LB. Intra-arrest cooling improves outcomes in a murine cardiac arrest model. Circulation. 2004; 109: 2786–2791.
Nozari A, Safar P, Stezoski SW, Wu X, Kostelnik S, Radovsky A, Tisherman S, Kochanek PM. Critical time window for intra-arrest cooling with cold saline flush in a dog model of cardiopulmonary resuscitation. Circulation. 2006; 113: 2690–2696.
Kim F, Olsufka M, Carlbom D, Deem S, Longstreth WT Jr, Hanrahan M, Maynard C, Copass MK, Cobb LA. Pilot study of rapid infusion of 2 L of 4°C normal saline for induction of mild hypothermia in hospitalized, comatose survivors of out-of-hospital cardiac arrest. Circulation. 2005; 112: 715–719.
The use of mild hypothermia (32°C to 24°C) in resuscitated ventricular fibrillation patients after hospital arrival has been shown to improve survival and neurological outcomes. Results from animal models of cardiac arrest, however, suggest that efficacy of mild hypothermia would improve if initiated as soon as possible after return of spontaneous circulation. Rapid cooling in the field by paramedics as soon as possible after return of spontaneous circulation has been hypothesized to increase survival and/or neurological outcome. In the present pilot study, we examined the safety, efficacy, and feasibility of using a rapid infusion of 4°C normal saline by paramedics in the field after return of spontaneous circulation in 125 patients who suffered cardiac arrest from ventricular fibrillation, asystole, or pulseless electrical activity. Sixty-three received a rapid infusion of up to 2 L cold normal saline, resulting in a mean temperature decrease of 1.24±1°C with a hospital arrival temperature of 34.7°C, whereas the 62 patients not randomized to cooling experienced a mean temperature increase of 0.10±0.94°C (P<0.0001) with a hospital arrival temperature of 35.7°C. In-field cooling was not associated with adverse consequences in terms of blood pressure, heart rate, arterial oxygenation, evidence for pulmonary edema on initial chest x-ray, or rearrest. Secondary end points of awakening and discharged alive from hospital trended toward improvement in ventricular fibrillation patients randomized to in-field cooling. We propose that the effect of this cooling method on neurological outcome after cardiac arrest be studied in larger numbers of patients.
Clinical trial registration information—URL: http://www.clinicaltrials.gov. Unique identifier: NCT00329563.