Use of Vasopressin

[Nate]

What a concept, eh?

Here's the "options" my favourite medical director gave us:

• 1. Do whatever you think the patient needs.

2. Don't call me until you have done it and are enroute to the ER.

OMG, you mean you can talk to the doctor? In Houston if you call a doctor you'll be at the ER before they pick up the phone, and you'll probably retire from EMS before they pick up the phone.

I just went from a service that carried the bare cardiac drugs to one that gives me extra drugs (not just cardiac), the ones that make being a paramedic easier because you can actually treat people. I'm like a kid in a candy shop at this new service.

Hold on, I think I have a drug for that. My old service didn't carry phenergran, now this service that I'm at does. All of my old patients always c/o N/V/D and would give me a little sample...here no one does.

[Ace844]

What a concept, eh?

Here's the "options" my favourite medical director gave us:

• 1. Do whatever you think the patient needs.

2. Don't call me until you have done it and are enroute to the ER.

My kind oif med. Director....This Md needs to spread the love...AROUND THE COUNTRY!!! We could certainly use more of those here!!

Ace

[Ace844]

Hi All,

I have a question somewhat on the subject of this thread...so please forgive me for digressing just a bit.

Does anyone's service and or practice environment allow them to start a Vasopressin infuison in sepsis/DKA as well???? just curious..

out here,

Ace844

[Nate]

My kind oif med. Director....This Md needs to spread the love...AROUND THE COUNTRY!!! We could certainly use more of those here!!

Ace

Thats how it is around here...then again most doctors don't want to be bothered around here unless it requires them to gain money. Calling a head and asking an ER to give you an order...doesn't happen around here unless he knows you personally.

[Ace844]

Thats how it is around here...then again most doctors don't want to be bothered around here unless it requires them to gain money. Calling a head and asking an ER to give you an order...doesn't happen around here unless he knows you personally.

Perhaps they don't have alot of confidence in the general population of EMS providers clinical ability in your area.

Most docs I have spoken with state that if you present your patient in a thorough, concise, professional manner be clear about the what/why of the intervention you are asking for...then you aren't likely to be denied. Being known as a professional with sound clinical judgement also enhances your chances that they will give you the go ahead and or do Med con. the way "Dustdevil" states he has it...It should be noted that this is my experience in my practice environment..YMMV....

Ace844

[Nate]

There are over 250 ambulance services that use Memorial Hermann Hospital in the Texas Medical Center...it is next to impossible to remember which one has that "bad apple" when EMS has such a high turn over rate.

So they generally won't give orders to someone they don't know. That is why I make it a point to know my patient, so that if I ever need to get an order the doctor remembers me.

Most of the orders I ask for are pedi and thats because I have some wild idea and I need a doctors okay to deviate from protocol. (Such as my atropine in an X OD that had the 13 year old talking to me by the time we got to the hospital.)

[Dustdevil]

Just to clarify, the medical director I was referring to didn't ask us not to call because he didn't trust us. Quite the contrary. He asked us not to call because he trusted us and expected us to make our own decisions regarding patient care, no matter what the situation might be.

[flight-lp]

There are over 250 ambulance services that use Memorial Hermann Hospital in the Texas Medical Center...it is next to impossible to remember which one has that "bad apple" when EMS has such a high turn over rate.

So they generally won't give orders to someone they don't know. That is why I make it a point to know my patient, so that if I ever need to get an order the doctor remembers me.

All of 'em have one or two bad apples!!!!! Since you Quoted MHHS-Hermann Hospital, I will offer another top reason why they are less than cooperative in giving pre-hospital orders, it is because most physicians that will actually talk to an EMS crew are residents and are not allowed to give on-line medical control orders per UT med school policy. Bottom line is these guys (and gals) are busy, they do not have time to consult with EMS. Hermann has a medical control officer on duty (a Paramedic in the LifeFlight office) who consults with the few agencies that are involved in an online medical control program. Most 911 providers in the area though are completely offline with standing orders, my service included. Usually the only reason why I transport to Hermann is for level I trauma cases, none of which should require any deviation from the basic trauma standards of care. Any medical patient that I have that would require such a deviation to save their life and that would require me to speak to a physician in the field would need to go to a closer facility. Houston is fortunate to have such a wide variety of services available in the outlying metropolitan areas, going "downtown to Hermann" isn't needed as often as it is used. The amount of time it takes to finally get through to a physician can also be used to thoroughly re-think and reassess your patient and then make an appropriate decision for their care. Just my .02 cents worth anyways.

[Ace844]

(Vasopressin or Epinephrine for Out-of-Hospital Cardiac Arrest

Peter C. Wyer @ MDa, Phillips Perera, MDa, Zhezhen Jin, PhDb, Qi Zhou, PhDd, Deborah J. Cook, MDcd, Stephen D. Walter, PhDd, Gordon H. Guyatt, MDcd

Received 13 January 2005; received in revised form 1 June 2005, 30 August 2005 and 1 November 2005; accepted 14 November 2005 published online 13 February 2006.)

Study objective

The use of vasopressin in patients with cardiac arrest presenting with specific rhythms is controversial. We performed an evidence-based emergency medicine review of evidence comparing vasopressin to epinephrine in structured cardiac arrest protocols.

Methods

We searched MEDLINE, EMBASE, the Cochrane Library, and other databases for randomized trials or systematic reviews comparing vasopressin to epinephrine for adults with cardiac arrest and measuring survival to hospital discharge and neurologic function in survivors. We used standard criteria to appraise the quality of published trials and systematic reviews. We used the random effects model in supplementary analyses to summarize results and to test for significant differences across subgroups of patients presenting with different arrest rhythms.

Results

We found 3 high-quality well-reported randomized trials and 1 rigorous meta-analysis. The evidence does not confirm a consistent benefit of vasopressin over epinephrine in increasing survival or improving neurologic outcome in survivors. Subgroup analysis reveals a large difference in effect of vasopressin over epinephrine in cardiac arrest patients with asystole, compared to other arrest rhythms, coming from within-trial comparisons. The difference is not consistent across otherwise similar trials, is not statistically significant, may reflect the application of multiple unplanned subgroup analyses, and is not supported by a plausible biological hypothesis.

Conclusion

Evidence from randomized trials does not establish a benefit of vasopressin over epinephrine in increasing survival to discharge or improving neurologic outcomes in adult patients with nontraumatic cardiac arrest.

a Emergency Medicine Residency Program, New York Presbyterian Hospital, New York, NY

b Department of Biostatistics, Columbia University, New York, NY

c Department of Medicine, McMaster University, Hamilton, Ontario, Canada.

d Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada.

[Ace844]

(Annals of Emergency Medicine

Volume 48 @ Issue 1 , July 2006, Pages 86-97

doi:10.1016/j.annemergmed.2005.11.024

Copyright © 2006 American College of Emergency Physicians Published by Mosby, Inc.

Cardiology/evidence-based emergency medicine review

Vasopressin or Epinephrine for Out-of-Hospital Cardiac Arrest

Peter C. Wyer MDa, , , Phillips Perera MDa, Zhezhen Jin PhDb, Qi Zhou PhDd, Deborah J. Cook MDc, d, Stephen D. Walter PhDd and Gordon H. Guyatt MDc, d

aEmergency Medicine Residency Program, New York Presbyterian Hospital, New York, NY

bDepartment of Biostatistics, Columbia University, New York, NY

cDepartment of Medicine, McMaster University, Hamilton, Ontario, Canada.

dDepartment of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada.

Received 13 January 2005; revised 1 June 2005, 30 August 2005, 1 November 2005; accepted 14 November 2005. Available online 10 February 2006.)

Study objective

The use of vasopressin in patients with cardiac arrest presenting with specific rhythms is controversial. We performed an evidence-based emergency medicine review of evidence comparing vasopressin to epinephrine in structured cardiac arrest protocols.

Methods

We searched MEDLINE, EMBASE, the Cochrane Library, and other databases for randomized trials or systematic reviews comparing vasopressin to epinephrine for adults with cardiac arrest and measuring survival to hospital discharge and neurologic function in survivors. We used standard criteria to appraise the quality of published trials and systematic reviews. We used the random effects model in supplementary analyses to summarize results and to test for significant differences across subgroups of patients presenting with different arrest rhythms.

Results

We found 3 high-quality well-reported randomized trials and 1 rigorous meta-analysis. The evidence does not confirm a consistent benefit of vasopressin over epinephrine in increasing survival or improving neurologic outcome in survivors. Subgroup analysis reveals a large difference in effect of vasopressin over epinephrine in cardiac arrest patients with asystole, compared to other arrest rhythms, coming from within-trial comparisons. The difference is not consistent across otherwise similar trials, is not statistically significant, may reflect the application of multiple unplanned subgroup analyses, and is not supported by a plausible biological hypothesis.

Conclusion

Evidence from randomized trials does not establish a benefit of vasopressin over epinephrine in increasing survival to discharge or improving neurologic outcomes in adult patients with nontraumatic cardiac arrest.

Article Outline

Clinical scenario

Formulating the question

Searching for and selecting the best evidence

Examining the evidence

Description of the trials

Primary results of the trials

Subgroup analysis

Was the subgroup difference suggested by comparisons within rather than between studies?

Was the magnitude of the subgroup difference large?

Was the subgroup difference consistent across studies?

Was the subgroup difference statistically significant?

Applying the evidence

Critically Appraised Topic (CAT): Does vasopressin in place of epinephrine improve survival to discharge without worsening neurological function in patients with out-of-hospital cardiac arrest?

References

Clinical scenario

You are the medical and education director of a regional emergency medical system with basic and advanced life support capabilities. Publicity about a recent trial of vasopressin as an alternative to epinephrine for patients with cardiac arrest leads several members of your training committee to ask when your system is shifting to that alternative. They point out that the American Heart Association guidelines for cardiac resuscitation that you use as the basis for your own advanced life support protocols already list vasopressin as an option. Your paramedic ambulances do not stock vasopressin, nor do your training and recertification programs include it. You decide to examine the evidence favoring vasopressin over epinephrine before making a major revision in your protocols.

The following evidence-based emergency medicine review1 seeks an answer to the question posed by this scenario.

Formulating the question

Hospital inpatients receiving cardiac resuscitation might have a prognostic advantage over patients outside of the hospital by virtue of earlier recognition and more rapid initiation of basic and advanced life support interventions. On the other hand, inpatients’ prognosis may be adversely affected by their concomitant acute conditions. The magnitude and direction of difference in effect of vasopressin on cardiac arrest outcomes between patients inside and outside of the hospital are unpredictable. We therefore included inpatient studies.

We confined our review to survival to hospital discharge and good neurologic function, outcomes that we believe patients themselves most value. Patients and families might consider admission to an ICU with no increased chance of survival to discharge to constitute an undesirable consequence of a new resuscitation drug. For expediency, resuscitation trialists sometimes define “survival to admission” as their primary outcome measure.2 and 3 Unfortunately, an intervention’s impact on a surrogate outcome such as survival to hospital admission is no guarantee of patient-important benefit.4 and 5 Several studies suggest that survival to ICU admission is not a good surrogate for survival to hospital discharge.2, 3 and 6

We formulated our question as: In patients with cardiac arrest not attributable to trauma or environmental exposures, what is the impact on survival to hospital discharge and survivor neurologic function of vasopressin compared to epinephrine, administered at the first point in a cardiac arrest protocol at which epinephrine would routinely be given?

Searching for and selecting the best evidence

We searched for randomized trials and systematic reviews comparing vasopressin to epinephrine for adults in cardiac arrest. We searched MEDLINE from 1966 to July 2004 and EMBASE from 1980 to January 2004 with the OVID interface, using search terms “vasopressin,” “epinephrine,” “cardiac arrest,” and “heart arrest,” with no language restrictions. We limited our MEDLINE search, but not our other searches, to randomized trials or systematic reviews using epinephrine as the comparison intervention. We included randomized trials or systematic reviews of randomized trials, with no other exclusion criteria. Our search yielded 271 results. We also searched all databases of the Cochrane Library7 through 2004 issue 1, Emergency Medical Abstracts (available online at http://ccme.org) from 1977 through December 2004,8 and online resources including BestBETS (available online at http://www.bestbets.org), using the single search term “vasopressin.” These databases yielded a total of 1,036 results. We reviewed the bibliographies of eligible trials and systematic reviews and of selected non–systematic reviews and commentaries for citations of additional eligible articles. Finally, we searched the bibliography of the relevant sections of the 2000 update of the American Heart Association Advanced Cardiac Life Support guideline.9

We found 3 trials comparing vasopressin to epinephrine in patients with cardiac arrest, 2 out-of-hospital10 and 11 and 1 limited to inpatients.12 One published systematic review,13 1 protocol in the Cochrane Database of Systematic Reviews,14 and 1 “shortcut review,”15 as well as our bibliographical reference review, revealed no additional trials. We found no placebo-controlled trials or other human randomized trials involving vasopressin for patients with cardiac arrest. The systematic review by Biondi-Zoccai et al13 was limited in its reporting, was not restricted to clinical trials, and did not include the most recent and largest randomized trial by Wenzel et al.11 After the completion of our primary searches and after the initial submission of our review, a second, well-reported, systematic review appeared that used inclusion criteria identical to our own.16 Aung and Htay16 identified the same 3 trials identified by our searches, as well as 2 additional clinical trials.17 and 18 We will focus the rest of this review on the Aung and Htay16 meta-analysis, supplemented by elements of our own analysis that either differ from or go beyond that of Aung and Htay.16

Examining the evidence

Aung and Htay16 restricted their review to randomized trials comparing vasopressin to epinephrine in patients with cardiac arrest and reporting patient-important outcomes. They independently selected studies for inclusion.

Aung and Htay16 report a rigorous search of MEDLINE, EMBASE, the Cochrane Library, CINAHL, bibliographies of related articles, registries of conference proceedings, and unpublished trials. They found a statistical test for evidence of studies not identified by their search to be negative. In addition to the 5 trials mentioned previously,10, 11, 12, 17 and 18 Aung and Htay16 also report the existence of a trial in progress, the Cardiac Arrest Research Project being conducted by the University of Pittsburgh (available online at http://newsbureau.upmc.com/emergency/vasopressin04.htm). This trial compares vasopressin to placebo as an addition to standard therapy including epinephrine and will not be eligible for inclusion in our review.

Description of the trials

Table 1 summarizes the key features of the 5 randomized trials identified by Aung and Htay16 and included in their analysis. Of the 2 trials not identified in our own independent search, one was limited to 10 patients and was published in abstract form only,17 and the other was published in a Chinese journal.18 The inclusion of non-English studies and abstracts in systematic reviews is controversial.19 and 20 Exclusion of unpublished studies may lead to underestimation of treatment effect.19 and 21 On the other hand, inclusion of non-English and incompletely reported studies may have little impact on the results of most systematic reviews.22 and 23

Table 1.

Summarizing the characteristics of 5 randomized trials comparing vasopressin to epinephrine in patients with cardiac arrest. Study Patients Interventions Comparisons Outcomes

Wenzel et al,11 2004 1,186 European adult patients with out-of-hospital arrest, average age 66 years, 70% men. Excluded terminally ill patients, those successfully defibrillated without drugs or with trauma; 61% arrests attributed to cardiac causes; 40% ventricular fibrillation, 44% asystole; 78% arrests witnessed, average time from arrest to basic life support 7.9 min and to advanced life support 14.9–15.6 min 40 IU vasopressin intravenously either immediately or after 3 attempts at defibrillation

Dose repeated in 3 minutes if no return of circulation

1 mg Of epinephrine intravenously after same protocol as for vasopressin Survival to hospital admission and to discharge, neurologic function by cerebral performance scale24 and 25

Stiell et al,12 2001 200 Adult patients with cardiac arrest in Canadian hospitals or emergency wards; 81% inpatients, 50% ward patients, 22% ICU. Average age 70 years, 64% men. Excluded terminally ill patients or those with trauma or exsanguination; 30% of arrests attributed to cardiac causes; 18% ventricular fibrillation, 31% asystole; 81% of arrests witnessed, average time from arrest to basic life support 1.4–1.9 min and to advanced life support 2.5–3.2 min 40 IU Vasopressin intravenously at the point in ACLS protocols that epinephrine first indicated.

One dose only

1 mg epinephrine intravenously after same protocol as for vasopressin Survival to hospital discharge, neurologic function by cerebral performance scale 25 and the Mini-Mental State examination38

Lindner et al,10 1997 40 Adult European patients with out-of-hospital ventricular fibrillation arrest. Average age 65 years, 72% men. Excluded patients with trauma or terminal illness and patients who received epinephrine by endotracheal tube.

Average time from arrest to basic life support 6.1–6.5 min, and to advanced life support 13.9–15.1 min

40 IU Vasopressin intravenously after shocks failed to restore rhythm.

One dose only

1 mg Epinephrine intravenously once after shocks failed to restore rhythm Survival to admission and to hospital discharge, neurologic function by Glasgow Coma Scale score

Lee et al, 200017 10 Patients in a large university teaching hospital. Other details ambiguous or not reported. 40 Units vasopressin given 0.1 unit/kg/min “as a bolus” Epinephrine, dose not stated Return of spontaneous circulation not otherwise described. Neurologic outcome not otherwise described.

Li et al, 199918 83 Adult Chinese hospital inpatients. Average age 57 years, 67% men. “Average cardiac arrest time” 10 min, not otherwise characterized. Other population details not reported. 0.5 U/kg Vasopressin or 1.0 U/kg vasopressin administered every 10 min 1.0 mg Epinephrine or 5.0 mg epinephrine administered every 5 minutes Hospital discharge.

No neurologic assessment reported.

ACLS, Advanced cardiac life support.

Two members of our own author team fluent in Chinese (QZ and ZJ) reviewed the full text of the study by Li et al.18 The reports by Lee17 and Li18 do not allow full assessment of their study populations, treatment protocols, or the susceptibility to bias, nor do they provide information about the distribution of presenting rhythms between their study groups. We are unclear about the relationship of the results reported by Lee et al17 to our own target outcomes. Aung and Htay16 report failed attempts to obtain more information from the authors of these 2 trials. We did not attempt to reproduce their inquiry. The inclusion of these studies in a pooled analysis might distort the estimates in ways that the parsimoniously reported data does not allow an investigator to anticipate or assess.

The 3 fully reported trials encompass a broad range of settings and variability within potentially important parameters, such as time from arrest to initiation of advanced life support. For example, in contrast to the patients studied by Wenzel et al11 and Lindner et al,10 the patients studied by Stiell et al12 had concomitant acute problems warranting emergency department or hospital admission. Twenty-two percent were in the ICU at arrest.12 Aung and Htay16 did not perceive these factors to prohibit pooling of the results of their primary outcomes across the trials, nor do we perceive them to constitute a priori incompatibility of these studies.

The 2 largest trials11 and 12 used a common measure of neurologic outcome, a previously published cerebral performance instrument characterized by clear and discrete gradations of functional recovery, ranging from death to full recovery.24 Other investigators have used this instrument in cardiac resuscitation research.25 A preliminary report by Nesbitt et al26 indicates that this instrument correlates well with a highly validated measure of health-related quality of life27 but may overestimate the extent of functional recovery in survivors.26

Aung and Htay16 performed independent assessment of the quality of the trials included in their review, with particular attention to concealment of randomization, blinding, completeness of follow-up, and outcome assessment.16 They found uniformly high quality in the 3 assessable trials using these criteria,10, 11 and 12 although the blinding of outcome assessment was not explicit in the study by Lindner et al.10 Aung and Htay16 report 80% agreement for their independent assessments (κ=.64).

Table 2 summarizes our own assessment of the likelihood of bias within the 3 fully reported trials, using criteria similar to those used by Aung and Htay.16 and 28 The loss of 76 patients to complete follow-up (6% of patients randomized) in the Wenzel et al11 study is potentially problematic. If these patients and their outcomes were not randomly distributed between the treatment groups, the results of the trial might be substantially affected.

Table 2.

Summarizing the assessment for susceptibility to important bias of the 3 fully reported trials. Criterion Wenzel et al,11 2004 Stiell et al,12 2001 Lindner et al,10 1997

Randomization Multicenter randomized trial. Randomization blocked in groups of 10 and stratified by center Multicenter randomized trial. Random distribution of study drugs to crash carts in treatment centers. Stratified by center Single-center randomized trial. Computer-generated randomization of identical syringes

Concealment Adequate Adequate Adequate

Intention to treat Yes 74 (27%) Patients were excluded post hoc by means of blinded adjudicated revision of eligibility assessment; 50 were due to cardiac arrest before arrival in hospital. The others were due to protocol violations or ineligibility caused by clinical circumstances Yes

Baseline comparisons 38% of patients receiving vasopressin presented with ventricular fibrillation versus 41% of patients receiving epinephrine 20% Of patients receiving vasopressin presented with ventricular fibrillation versus 16% of patients receiving epinephrine. Treatment times were somewhat longer in vasopressin group Well balanced within limits of small number of allocations

Blinding Blinded with respect to patients, care providers, and data collectors, except for 5 patients whose protocols were broken after hospital admission. Reporting of blinding not explicit for assessors of neurologic outcome or data analysts Blinded with respect to patients, care providers, and data collectors. Reporting of blinding not explicit for assessors of neurologic outcome or data analysts Blinded with respect to patients, care providers. Reporting of blinding not explicit for data collectors, assessors of neurologic outcome, or data analysts

Cointerventions Bystander CPR 19% vasopressin, 18% epinephrine. Time from BLS to ACLS interventions 7.0 min in vasopressin group, 7.7 min in epinephrine group. About 2% more of the vasopressin group received lytics, and 2% fewer received amiodarone and atropine compared to epinephrine group Time from arrest to study drug 3.2 min vasopressin, 2.5 min epinephrine. No consistent pattern of imbalance in reported drug therapies other than the study drugs Bystander CPR 20% vasopressin, 25% epinephrine. Time from arrest to study drug 15.1 min vasopressin, 13.9 min vasopressin. Distribution of other cointerventions not reported

Complete follow-up 33 (2.8%) Patients could not be included in analysis because of missing study-drug codes. Additionally, 20 patients, equally divided, were lost to follow-up before hospital discharge, and another 23 patients who survived to discharge were lost to neurologic follow-up. Complete Complete

CPR, Cardiopulmonary resuscitation; BLS, basic life support.

Primary results of the trials

Aung and Htay16 chose the random effects model to summarize data from the 5 trials they included.29 and 30 The random effects model usually leads to wider confidence intervals (CIs) around the pooled result than does the fixed effects model and may be considered “more conservative” for this reason.31 Aung and Htay16 used standard methods to assess for statistical heterogeneity between the results of the included studies. They used the I2 statistic to provide an estimate of the percentage of the variability between the results of studies being pooled that is due to true differences between the studies, as opposed to variability due to chance alone.32 and 33 When all 5 trials were included, Aung and Htay16 report a high degree of heterogeneity for the primary outcome of death before hospital discharge.16 The χ2 test for heterogeneity yielded a P value of .09, which is less than the commonly preferred cutoff of .1. I2 was 34%, suggesting a substantial portion of variability between the studies to be due to actual differences.

In our own analysis, using odds ratios for the same outcome and including only the 3 fully reported trials, the P value for heterogeneity was .21, and the I2 value was 35%. Most of the heterogeneity is due to a trend in the direction of benefit of vasopressin in the study by Lindner et al,10 a trend not observed in the studies by Wenzel et al11 and Stiell et al.12 and 16 A review of the characteristics of the Lindner et al10 study, compared to the analogous characteristics of the studies by Wenzel et al11 and Stiell et al,12 (Table 1) fails to reveal consistent differences that would explain this difference in trend. Under the circumstances, pooling of these 3 studies, as elected by Aung and Htay,16 may reasonably provide a more generalizable result.34

Aung and Htay16 report a pooled relative risk of death before hospital discharge, vasopressin compared to epinephrine, across the 3 fully reported trials of 0.99 (95% CI 0.95 to 1.02). When the trials by Li et al18 and Lee et al17 were included, the relative risk was 0.96 (95% CI 0.87 to 1.05). Our own analysis also used the random effects model and preferred odds ratios to risk ratios because of important inconsistencies between the results when risk ratios were used. Such inconsistencies commonly occur when outcome rates are extremely high, as in cardiac arrest. Our pooled odds ratio for the outcome of predischarge mortality, vasopressin compared to epinephrine, was 0.91 (95% CI 0.52 to 1.57). In all analyses, the CI around the pooled effect included values favoring epinephrine.

Aung and Htay16 analyzed the effect of vasopressin on death before hospital discharge or neurologic impairment. They grouped patients with only moderate disability together with those who died or were in a persistent vegetative states and included the study by Lee et al17 with those by Stiell et al12 and Wenzel et al.11 The Lee et al17 study contributed substantial heterogeneity to this analysis (I2 of 34%).16 Aung and Htay16 report a pooled relative risk of death or neurologic impairment, vasopressin compared to epinephrine, of 1.0 (95% CI 0.94 to 1.07).

A disadvantage of the Aung and Htay16 approach to assessment of neurologic outcome is that a patient with some residual neurologic deficit but still able to engage in part-time employment is grouped with patients who die or survive with major functional or cognitive impairment. To correct for this disadvantage, including data from the 2 trials that used a common measure of outcome, we classified patients in the 2 worst neurologic outcome categories by the cerebral performance instrument 24 and 25 used by Stiell et al12 and Wenzel et al11 as “poor neurologic outcome.” Such patients may have some cognitive function but no independence in activities of daily living. A patient without poor neurologic function by these criteria is at least able to engage in part-time employment in a sheltered environment.25 We believe that this definition of the outcome is more likely to cohere with the values of patients and their families than is that used by Aung and Htay.16

Using this revised composite outcome of death or major disability, the pooled odds ratio, vasopressin compared to epinephrine, is 1.32, 95% CI 0.82 to 2.14 (Figure 1). We found no important trends with respect to other measures of cognitive function used by Stiell et al12 and Lindner et al.10

(38K)

Figure 1. Forest plot of outcome of death or severe disability in the 2 trials using a common measure of neurologic outcome. For each trial, the small square corresponds to the observed odds ratio for predischarge mortality, and the horizontal line defines the 95% CI. An odds ratio of 1, identified by the vertical line, would reflect an identical effect of the 2 drugs. The lowest plot provides the pooled random effects odds ratio and CI. Poor neurologic outcome is defined as a score of 3 or greater on the cerebral performance score.25

Subgroup analysis

Wenzel et al11 emphasized an apparent positive effect of vasopressin in decreasing hospital mortality among patients presenting with asystole when advanced life support interventions were initiated. Three trials report data on predischarge mortality in subgroups of patients defined by presenting rhythm at initiation of resuscitation.10, 11 and 12 Lindner et al’s10 small trial was confined to patients presenting with ventricular dysrhythmias.

Wenzel et al11 also reported apparent benefit of initial vasopressin over initial epinephrine among patients receiving additional doses of epinephrine after the 2 doses of the study drug provided for in the protocol failed to result in return of spontaneous circulation. Such patients are not prospectively identifiable, nor has this effect been reported in other trials.

Aung and Htay16 detected no statistically significant difference between subgroups defined by presenting rhythm. Their analysis does not, however, exhaust the issue raised by Wenzel et al11 and others. We conducted a systematic subgroup analysis using published validity criteria (Table 3).35 and 36 We confined our consideration of the effect of vasopressin in subgroups to the hypothesis that vasopressin is of benefit in patients with asystole but not in those with ventricular fibrillation. We present the criteria in suggested order of application.

Table 3.

Summary of appraisal of presenting rhythm subgroups in 3 randomized trials comparing vasopressin to epinephrine in patients with cardiac arrest based on criteria proposed for evaluation of subgroup analyses in randomized trials.35 and 36 The hypothesis of a selective benefit of vasopressin over epinephrine in patients resenting with asystole fails with respect to several criteria. Criterion Conclusion

Was the subgroup difference suggested by comparisons within rather than between studies? Yes. The effect of vasopressin compared to epinephrine on predischarge mortality was reported for the asystole subgroup within 2 of the 3 trials.

Was the magnitude of subgroup difference large? Yes. The pooled odds ratio of 0.43 for predischarge mortality for patients in asystole must be considered a large effect compared to the pooled odds ratio of 1.00 for patients in the ventricular fibrillation subgroup (Figure 3).

Was the subgroup difference consistent across studies? No. Only 1 trial showed benefit in patients presenting with asystole. One trial showed a trend toward benefit in patients presenting with ventricular fibrillation, a trend that was not observed in the other 2 trials.

Was the subgroup difference statistically significant? No. The effect of vasopressin compared to epinephrine on predischarge mortality in patients with asystole was not statistically different from that in patients with ventricular fibrillation.

Did the trialists plan the subgroup analysis in advance? Unclear. The only study to report a subgroup effect did not report advance planning of the analysis.

Were many subgroup analyses performed and selectively reported? Unclear in the Wenzel et al 11 trial. Stiell et al 12 considered many subgroups.

Is the difference in effect in the subgroup supported by biological hypothesis? No. A biological hypothesis supporting a selective benefit of vasopressin in patients presenting in asystole or in another rhythm class has not been elaborated in either the trials or other sources reviewed in preparation for our analysis.

Was the subgroup difference suggested by comparisons within rather than between studies?

When investigators conduct independent trials on different subgroups of patients (eg, one trial includes only patients with asystole, another only patients with ventricular fibrillation), apparent differences in effect between the studies may originate from differences in the study populations other than that hypothesized (differences in comorbidity rather than differences in cardiac rhythm) or in aspects of study design (blinding of clinicians or outcome assessors) rather than from differences in response between the subgroups on whom the investigators focus (ventricular fibrillation versus asystole). The apparently greater benefit of vasopressin in patients with asystole reported by Wenzel et al11 is an example of a “within-study” comparison (both Wenzel et al11 and Stiell et al12 included patients with ventricular fibrillation and asystole), and this strengthens the hypothesis that the difference in effect may be real.

Was the magnitude of the subgroup difference large?

The larger the observed difference in effect between subgroups, the less likely it is to have arisen by chance alone. SDW and QZ performed an exact analysis using a logistic regression model to assess the magnitude and precision of subgroup effects. We believe this method to be more reliable in the setting at hand because exact models are generally preferable, and particularly so when, as here, the data are sparse.

The pooled odds ratio for predischarge mortality of patients treated with vasopressin compared to those treated with epinephrine from our exact inference analysis was 0.43 for the asystole subgroups reported by Wenzel et al11 and Stiell et al,12 contrasting with 1.00 for the ventricular fibrillation subgroups of all 3 trials. This large difference supports the subgroup hypothesis.

Was the subgroup difference consistent across studies?

Wenzel et al11 observed a trend in the direction favoring vasopressin over epinephrine among patients presenting with asystole and a trend in the opposite direction among patients presenting with ventricular fibrillation (and with pulseless electrical activity). In the trial by Stiell et al,12 the trends in both asystole and ventricular fibrillation subgroups favored epinephrine. The Lindner et al10 study of patients with ventricular fibrillation observed a trend in favor of vasopressin. Hence, a consistent pattern of subgroup effects does not emerge from the trials to date.

Was the subgroup difference statistically significant?

Investigators and authors sometimes contrast a statistically significant difference between treatment and control in one subgroup (such as those with asystole) with the lack of statistical significance in another subgroup (such as those with other presenting rhythms). This, however, misses the important question, can the difference between the apparent effects in different subgroups (asystole versus ventricular fibrillation) be explained by chance?36

Aung and Htay16 applied a statistical test for heterogeneity to pooled results in each of the 3 rhythm subgroups. This approach is illustrated in simplified form in Figure 2. In both the Aung and Htay16 analysis and our own, the tests for significance yield P values well above .1 and low values of I2, indicating that the observed effect of vasopressin compared to epinephrine in patients with asystole is consistent with a hypothesis of a uniform underlying effect across all rhythm subgroups. This falls short of a direct test of the difference in comparative effect of vasopressin between asystole and other rhythm subgroups. Our own analysis illustrates such a direct test.

(39K)

Figure 2. Forest plot illustrating an indirect approach to testing for statistical significance of differences in subgroup effect. See legend to Figure 1 for the explanation of the plot. Subgroup data are pooled across the 2 studies that included patients with all 3 presenting rhythm subtypes. A statistical test for heterogeneity is applied. The P value of .51 indicates that the effect of vasopressin compared to epinephrine on mortality within these 3 subgroups is consistent with an underlying effect. The I2 value of 0 further suggests that all of the observed variation between the subgroups is attributable to chance.

We started our exact analysis by considering the possibility of an interaction between presenting rhythm subgroup and treatment effect. When we found no significant interaction, we repeated the analysis without the interaction term and computed exact tests of the subgroup and treatment main effects. The method was applied to all 3 trials (Figure 3)10, 11 and 12 and—because this is restricted to within-study comparisons of ventricular fibrillation and asystole subgroups—to the studies by Wenzel et al11 and Stiell et al.12

(55K)

Figure 3. Forest plot illustrating the pooled subgroup effects from the analysis. The analysis uses paired and unpaired data pertaining to the asystole and ventricular fibrillation subgroups from the trials by Wenzel et al,11 Stiell et al,12 and Lindner et al.10 See legend to Figure 1 for the explanation of the plot. Odds ratios for the effect on mortality of vasopressin compared to epinephrine are pooled within each subgroup. See text for further explanation.

Figure 3 displays the results of our statistical analysis of the subgroup effect. The question we are asking is, can the difference in the odds ratios in the asystole group (0.43) and the ventricular fibrillation group (1.00) be explained by chance? The ratio of these 2 odds ratios is 0.43, and the 95% CI is 0.12 to 1.37. The CI includes 1 and indicates that the difference between the original odds ratios of 0.43 and 1.00 is compatible with chance, ie, it is not statistically significant (P=.18).

The odds ratio for mortality for these 2 subgroups from the 3 trials is 0.87 (95% CI 0.58 to 1.29) favoring vasopressin. As a further control for study effect in the exact model, the ratio of odds ratios of effect on mortality of vasopressin in asystole compared to ventricular fibrillation subgroups in the strictly paired analysis is 0.38 (95% CI 0.11 to 1.22; P=.12), and the odds ratio for mortality in both subgroups combined is 0.95 (95% CI 0.62 to 1.44), both reflecting trends favoring vasopressin.

In summary, when a direct test of statistical significance is applied, the trend toward a mortality benefit of vasopressin compared to epinephrine in patients with asystole is not significantly different from the comparative effect in patients with ventricular fibrillation.

As summarized in Table 3, the remaining criteria for believability of the subgroup hypothesis about vasopressin compared to epinephrine in patients with asystole were not met.

Applying the evidence

Returning to our clinical scenario, as the director of emergency medical services and personnel in your region, you must consider a number of issues in deciding whether to upgrade the status of vasopressin in your protocols for patients in cardiac arrest. The recent trial of vasopressin compared to epinephrine received high publicity in the lay press. An editorial accompanying the Wenzel et al11 trial report in the New England Journal of Medicine called for unscheduled conventions of the American Heart Association and the American College of Cardiology to incorporate a recommendation of vasopressin for patients presenting in asystole.37 You consequently may be under some pressure to provide the perceived benefits of the new therapy to the region and may even be identified as the cause of any delay.

Changing regional emergency medical services protocols, however, entails considerable effort and expense, whether or not unscheduled meetings and conferences are required. Your regional emergency medical services committee would have to approve the new protocol, and you would have to administer a training update to all relevant care providers. To justify this effort, you need convincing evidence that a patient-important benefit of vasopressin exists.

We have summarized the data from a well-done meta-analysis and from the 3 fully reported randomized trials and have found no overall effect of vasopressin compared to epinephrine in reducing mortality before discharge. After systematically applying 7 published criteria for evaluation of subgroup analyses in randomized trials of effectiveness, we have concluded that the evidence of an important survival benefit of vasopressin over epinephrine for patients with asystole is not compelling. It also remains unclear whether vasopressin compared to epinephrine improves or worsens neurologic outcomes in survivors.

Our review is subject to the limitations inherent in shortcut reviews.1 However, a well-done meta-analysis and our own independent analysis all suggest that you may reasonably decide not to change cardiac arrest protocols until new evidence becomes available.

References

1 P.C. Wyer, B.H. Rowe and G.H. Guyatt et al., The clinician and the medical literature when can we take a shortcut?, Ann Emerg Med 36 (2000), pp. 149–155. Abstract | Abstract + References | PDF (38 K)

2 P. Dorian, D. Cass and B. Schwartz et al., Amiodarone as compared with lidocaine for shock-resistant ventricular fibrillation, N Engl J Med 346 (2002), pp. 884–890. Abstract-MEDLINE | Abstract-EMBASE | Abstract-Elsevier BIOBASE | Full Text via CrossRef

3 P.J. Kudenchuk, L.A. Cobb and M.K. Copass et al., Amiodarone for resuscitation after out-of-hospital cardiac arrest due to ventricular fibrillation, N Engl J Med 341 (1999), pp. 871–878. Abstract-EMBASE | Abstract-MEDLINE | Abstract-Elsevier BIOBASE | Full Text via CrossRef

4 G. Guyatt, V. Montori and P.J. Devereaux et al., Patients at the center in our practice, and in our use of language, ACP J Club 140 (2004), pp. A11–A12. Abstract-MEDLINE

5 H. Bucher, G. Guyatt and D. Cook et al., Therapy and applying the results surrogate outcomes. In: G. Guyatt and D. Rennie, Editors, Users’ Guides to the Medical Literature A Manual for Evidence-Based Clinical Practice, American Medical Association, Chicago, IL (2002), pp. 393–413.

6 C. Vandycke and P. Martens, High dose versus standard dose epinephrine in cardiac arrest a meta-analysis, Resuscitation 45 (2000), pp. 161–166. Abstract

7 Wiley InterScience. The Cochrane library. Available at: http://www3.interscience.wiley.com/cgi-bin.../106568753/HOME. Accessed June 1, 2005.

8 The Center for Medical Education. Emergency medical abstracts. Available at: http://ccme.org. Accessed June 1, 2005.

9 American Heart Association, Guidelines 2000 for cardiopulmonary resuscitation and emergency cardiovascular care, part 6 advanced cardiovascular life support, Circulation 102 (2000) (suppl), pp. I86–I166.

10 K.H. Lindner, B. Dirks and H.U. Strohmenger et al., Randomised comparison of epinephrine and vasopressin in patients with out-of-hospital ventricular fibrillation, Lancet 349 (1997), pp. 535–537. SummaryPlus | Full Text + Links | PDF (389 K)

11 V. Wenzel, A.C. Krismer and H.R. Arntz et al., A comparison of vasopressin and epinephrine for out-of-hospital cardiopulmonary resuscitation, N Engl J Med 350 (2004), pp. 105–113. Abstract-MEDLINE | Abstract-Elsevier BIOBASE | Full Text via CrossRef

12 I.G. Stiell, P.C. Hebert and G.A. Wells et al., Vasopressin versus epinephrine for inhospital cardiac arrest a randomised controlled trial, Lancet 358 (2001), pp. 105–109. SummaryPlus | Full Text + Links | PDF (86 K)

13 G.G.L. Biondi-Zoccai, A. Abbate and Q. Parisi et al., Is vasopressin superior to adrenaline or placebo in the management of cardiac arrest? a meta-analysis, Resuscitation 59 (2003), pp. 221–224. Abstract

14 I. Jacobs and F. Finn, Adrenaline and Vasopressin for Cardiac Arrest (Protocol) The Cochrane Library, John Wiley & Sons, Chichester, UK (2004).

15 Hogg K. Vasopressin or adrenaline in cardiac resuscitation [bestBETs Web site]. Available at: http://www.bestbets.org/cgi-bin/bets.pl?record=00407. Accessed March 2, 2004.

16 K. Aung and T. Htay, Vasopressin for cardiac arrest, Arch Intern Med 165 (2005), pp. 17–24. Abstract-MEDLINE | Abstract-EMBASE | Abstract-Elsevier BIOBASE | Full Text via CrossRef

17 C.C. Lee, Y.S. Jung and S.K. Yoon et al., Vasopressin administration in out-of-hospital cardiac arrest [abstract], Ann Emerg Med 36 (2000), p. S91.

18 P.J. Li, T.T. Chen and J.M. Zhang et al., Clinical study on administration of vasopressin during closed chest cardiopulmonary resuscitation, Chinese Crit Care Med 11 (1999), pp. 28–31.

19 S. Hopewell, S. McDonald and M. Clarke et al., Grey literature in meta-analyses of randomized trials of health care interventions, The Cochrane Database of Methodology Reviews (2002) Issue 4.

20 D.J. Cook, G.H. Guyatt and G. Ryan et al., Should unpublished data be included in meta-analyses?, JAMA 269 (1993), pp. 2749–2753. Abstract-MEDLINE | Abstract-EMBASE

21 L. McAuley, B. Pham and P. Tugwell et al., Does the inclusion of grey literature influence estimates of intervention effectiveness reported in meta-analyses?, Lancet 356 (2000), pp. 1228–1231. SummaryPlus | Full Text + Links | PDF (78 K)

22 P. Juni, F. Holenstein and J. Sterne et al., Does the inclusion of grey literature influence estimates of intervention effectiveness reported in meta-analyses?, Int J Epidemiol 31 (2002), pp. 115–123. Abstract-MEDLINE | Full Text via CrossRef

23 D. Fergusson, A. Laupacis and L.R. Salmi et al., What should be included in meta-analyses? an exploration of methodological issues using the ISPOT meta-analyses, Int J Technol Assess Health Care 16 (2000), pp. 1109–1119. Abstract-MEDLINE | Abstract-EMBASE | Full Text via CrossRef

24 B. Jennett and M. Bond, Assessment of outcome after severe brain damage a practical scale, Lancet 1 (1973), pp. 480–484.

25 The Brain Resuscitation Clinical Trial II Study Group, A randomized clinical trial of calcium entry blocker administration to comatose survivors of cardiac arrest design, methods, and patient characteristics, Control Clin Trials 12 (1991), pp. 525–545.

26 L.P. Nesbitt, I.G. Stiell and D. Cousineau et al., Is the cerebral performance category score a valid measure of functional outcome after out-of-hospital cardiac arrest?, Acad Emerg Med 12 (2005), p. 71 [abstract].

27 D. Feeny, K. Farris and I. Cote et al., A cohort study found the RAND-12 and Health Utilities Index Mark 3 demonstrated construct validity in high-risk primary care patients, J Clin Epidemiol 58 (2005), pp. 138–141. SummaryPlus | Full Text + Links | PDF (160 K)

28 G. Guyatt, D. Cook and P.J. Devereaux et al., Therapy. In: G. Guyatt and D. Rennie, Editors, Users’ Guides to the Medical Literature A Manual of Evidence-Based Clinical Practice, American Medical Association, Chicago, IL (2002), pp. 55–79.

29 J.L. Fleiss, The statistical basis of meta-analysis, Stat Methods Med Res 2 (1993), pp. 121–145. Abstract-MEDLINE

30 R. DerSimonian and N. Laird, Meta-analysis in clinical trials, Control Clin Trials 7 (1986), pp. 177–188. Abstract

31 V. Montori, G. Guyatt and A. Oxman et al., Summarizing the evidence fixed-effects and random-effects models. In: G. Guyatt and D. Rennie, Editors, Users’ Guides to the Medical Literature A Manual for Evidence-Based Clinical Practice, American Medical Association, Chicago, IL (2002), pp. 539–545.

32 J.P.T. Higgins and S.G. Thompson, Quantifying heterogeneity in a meta-analysis, Stat Med 21 (2002), pp. 1539–1558. Abstract-EMBASE | Abstract-Elsevier BIOBASE | Abstract-MEDLINE | Full Text via CrossRef

33 J.P.T. Higgins, S.G. Thompson and J.J. Deeks et al., Measuring inconsistency in meta-analyses, BMJ 327 (2003), pp. 557–560. Abstract-MEDLINE | Abstract-EMBASE | Abstract-Elsevier BIOBASE | Full Text via CrossRef

34 V. Montori, R. Hatala and G. Guyatt, Summarizing the evidence evaluating differences in study results. In: G. Guyatt and D. Rennie, Editors, Users’ Guides to the Medical Literature A Manual for Evidence-Based Clinical Practice, American Medical Association, Chicago, IL (2002), pp. 547–552.

35 A.D. Oxman and G.H. Guyatt, A consumer’s guide to subgroup analyses, Ann Intern Med 116 (1992), pp. 78–84. Abstract-MEDLINE | Abstract-EMBASE

36 A. Oxman and G. Guyatt, Summarizing the evidence when to believe a subgroup analysis. In: G. Guyatt and D. Rennie, Editors, Users’ Guides to the Medical Literature A Manual for Evidence-Based Clinical Practice, American Medical Association, Chicago, IL (2002), pp. 553–565.

37 K.M. McIntyre, Vasopressin in asystolic cardiac arrest, N Engl J Med 350 (2004), pp. 179–181. Abstract-MEDLINE | Abstract-Elsevier BIOBASE | Full Text via CrossRef

38 E.L. Teng and H.C. Chui, The modified Mini-Mental State (3MS) examination, J Clin Psychiatry 48 (1987), pp. 314–318. Abstract-MEDLINE

Critically Appraised Topic (CAT): Does vasopressin in place of epinephrine improve survival to discharge without worsening neurological function in patients with out-of-hospital cardiac arrest?

Question In patients with cardiac arrest not attributable to trauma or environmental exposures, what is the impact on survival to hospital discharge and survivor neurological function of vasopressin compared to epinephrine when administered at the first point in a cardiac arrest protocol at which epinephrine would routinely be given?

Reviewed by Wyer PC, Perera P, Jin Z, Zhou Q, Cook DJ, Walter SD, Guyatt GH

Date November 1, 2005

Expiration date November 1 2007

Clinical bottom line Vasopressin, administered to patients with out-of-hospital or in-hospital cardiac arrest, has no proven survival benefit compared to epinephrine when given at the same point in a structured resuscitation protocol, despite a possible increase in likelihood of hospital admission. A subgroup analysis of data from 3 randomized trials does not reveal a statistically significant benefit of vasopression in patients presenting with asystole or with other specific arrest rhythms. The trials suggest a possible trend towards worse neurological functional outcomes with vasopressin. Current evidence from randomized trials does not support vasopressin use in victims of cardiac arrest. A well done meta-analysis came to the same conclusion.

Search Strategy The search for randomized trials enrolling adults with cardiac arrest included MEDLINE, EMBASE, the Cochrane Library from the dates of origin through July of 2004, and Emergency Medical Abstracts from 1977 to December 2004. The MEDLINE search was limited to trials comparing vasopressin to epinephrine. Authors of a recent well done meta-analysis conducted an exhaustive search including registries of conference proceedings and unpublished trials.

Citations Primary: Meta-analysis

Aung K, Htay T. Vasopressin for cardiac arrest. Arch Intern Med. 2005;165:17-24.

Secondary: 3 fully reported trials

1. Wenzel V, Krismer AC, Arntz HR, et al. A comparison of vasopressin and epinephrine for out-of-hospital cardiopulmonary resuscitation. N Engl J Med. 2004;350:105-113.

2. Stiell IG, Hebert PC, Wells GA, et al. Vasopressin versus epinephrine for inhospital cardiac arrest: A randomised controlled trial. Lancet. 2001;358:105-109.

3. Lindner KH, Dirks B, Strohmenger HU, Prengel AW, Lindner IM, Lurie KG. Randomised comparison of epinephrine and vasopressin in patients with out-of-hospital ventricular fibrillation. Lancet. 1997;349:535-537.

Primary study characteristics Study population

Adult patients in Canada, Europe and Asia with cardiac arrest in and out of hospital. One study limited to patients with ventricular arrest rhythms. Settings and populations poorly described in 2 Asian studies.

Interventions

40 IU vasopressin or 1 mg epinephrine iv either immediately or after 3 attempts at defibrillation in patients with ventricular fibrillation. Dose repeated in 3 minutes if no return of pulse in the largest European study. Study protocol poorly described in 2 Asian studies.

Outcome Measures

Survival to hospital discharge in all studies. Neurological function by cerebral performance scale in survivors to discharge in 2 studies.

Study Design

Limited to randomized trials comparing vasopressin to epinephrine in human subjects.

Critical appraisal The systematic review employed an exhaustive search strategy, included registries of unpublished trials and abstracts and otherwise controlled for publication and selection bias. Quality appraisal and data abstraction were performed independently by 2 reviewers with substantial agreement above chance. Heterogeneity was assessed and was substantial when the 2 Asian studies were included. High quality was observed in 3 fully reported trials and could not be assessed in 2 Asian studies.

Results Primary and secondary outcomes

All trials reporting outcome Fully reported trials only

Pre-discharge mortality (Aung) RR+ .96 (.87, 1.05) RR+ .99 (.95, 1.02)

Death or any disability (Aung) δ RR+ 1.00 (.94, 1.07) —

Death or major disability δδ — OR+ 1.32 (.82, 2.14)

+ Relative risk (RR) or odds ratio(OR) and 95% confidence intervals vasopressin compared to epinephrine. ed to epinephrine. Values of RR or OR <1 favor vasopressin; values of RR or OR >1 favor epinephrine.

δ By cerebral performance score or by estimate of outcome from incompletely reported study.

δδ Data from trials of Wenzel and Stiell using the cerebral performance score. Patients in the lowest 2 categories were considered to have poor neurological outcome and were characterized as dependent on others for activities of daily living and with severe memory disturbance or in a vegetative state.

Subgroup analyses

Evidence supporting the hypothesis of selective benefit of vasopressin over epinephrine in cardiac arrest patients with asystole, compared to other arrest rhythms, includes a large difference in effect size coming from within-trial comparisons. The difference in effect of vasopressin in presenting rhythm subgroups, however, is not consistent across otherwise similar trials, is not statistically significant when appropriate analytical methods are applied, may reflect the application of multiple unplanned subgroup analyses and is not supported by a plausible biological hypothesis.

Supervising editor: Brian H. Rowe, MD, MScFunding and support: The authors report this study did not receive any outside funding or support.Reprints not available from the authors.

Address for correspondence: Peter C. Wyer, MD, Columbia University College of Physicians and Surgeons, Emergency Services, 446 Pelhamdale Avenue, Pelham Manor, NY 10803; 914-738-9368, fax 914-738-6537