Association of Vasopressin Plus Catecholamine Vasopressors vs Catecholamines Alone With Atrial Fibrillation in Patients With Distributive Shock

Key Points

Question

In patients with distributive shock (a condition due to excessive vasodilation, most frequently from severe infection), is the addition of vasopressin to catecholamine vasopressors superior to catecholamine vasopressors alone for atrial fibrillation?

Findings

In this systematic review and meta-analysis of 23 trials that included 3088 patients with distributive shock, the addition of vasopressin to catecholamine vasopressors compared with catecholamine vasopressors alone was significantly associated with a lower risk of atrial fibrillation (relative risk, 0.77).

Meaning

Addition of vasopressin to catecholamines may offer a clinical advantage for prevention of atrial fibrillation.

Abstract

Importance

Vasopressin is an alternative to catecholamine vasopressors for patients with distributive shock—a condition due to excessive vasodilation, most frequently from severe infection. Blood pressure support with a noncatecholamine vasopressor may reduce stimulation of adrenergic receptors and decrease myocardial oxygen demand. Atrial fibrillation is common with catecholamines and is associated with adverse events, including mortality and increased length of stay (LOS).

Objectives

To determine whether treatment with vasopressin + catecholamine vasopressors compared with catecholamine vasopressors alone was associated with reductions in the risk of adverse events.

Data Sources

MEDLINE, EMBASE, and CENTRAL were searched from inception to February 2018. Experts were asked and meta-registries searched to identify ongoing trials.

Study Selection

Pairs of reviewers identified randomized clinical trials comparing vasopressin in combination with catecholamine vasopressors to catecholamines alone for patients with distributive shock.

Data Extraction and Synthesis

Two reviewers abstracted data independently. A random-effects model was used to combine data.

Main Outcomes and Measures

The primary outcome was atrial fibrillation. Other outcomes included mortality, requirement for renal replacement therapy (RRT), myocardial injury, ventricular arrhythmia, stroke, and LOS in the intensive care unit and hospital. Measures of association are reported as risk ratios (RRs) for clinical outcomes and mean differences for LOS.

Results

Twenty-three randomized clinical trials were identified (3088 patients; mean age, 61.1 years [14.2]; women, 45.3%). High-quality evidence supported a lower risk of atrial fibrillation associated with vasopressin treatment (RR, 0.77 [95% CI, 0.67 to 0.88]; risk difference [RD], −0.06 [95% CI, −0.13 to 0.01]). For mortality, the overall RR estimate was 0.89 (95% CI, 0.82 to 0.97; RD, −0.04 [95% CI, −0.07 to 0.00]); however, when limited to trials at low risk of bias, the RR estimate was 0.96 (95% CI, 0.84 to 1.11). The overall RR estimate for RRT was 0.74 (95% CI, 0.51 to 1.08; RD, −0.07 [95% CI, −0.12 to −0.01]). However, in an analysis limited to trials at low risk of bias, RR was 0.70 (95% CI, 0.53 to 0.92, P for interaction = .77). There were no significant differences in the pooled risks for other outcomes.

Conclusions and Relevance

In this systematic review and meta-analysis, the addition of vasopressin to catecholamine vasopressors compared with catecholamines alone was associated with a lower risk of atrial fibrillation. Findings for secondary outcomes varied.

This meta-analysis compares the effects of vasopressin with vs without catecholamine vasopressors on atrial fibrillation, mortality, stroke, and other adverse outcomes in patients with distributive shock.

Introduction

In distributive shock, widespread vasodilation leads to decreased systemic vascular resistances and mean arterial pressure (MAP).¹ If not reversed, end-organ hypoperfusion results in significant morbidity; mortality rates reached 50% in observational studies conducted in 2013 and 2014.^2,3 Sepsis is the most common cause of distributive shock. It can also occur after cardiovascular surgery, spinal cord injury, or arise as a consequence of anaphylaxis or prolonged hypoperfusion.^1,4

Managing distributive shock involves treating the underlying cause, volume resuscitation, and infusing vasopressors to maintain a perfusing blood pressure.^5,6 Clinicians frequently use catecholaminergic vasopressors (eg, norepinephrine, epinephrine, dopamine, dobutamine). However, catecholamines have adverse effects including myocardial ischemia and arrhythmia,^6,7,8 which may affect outcomes.⁹

Atrial fibrillation is a common adverse event in patients with distributive shock and is independently associated with morbidity, mortality, and increases in length of stay (LOS).^10,11,12

Vasopressin, an endogenous peptide hormone, can also be used as a vasopressor. Patients with septic shock have relative vasopressin deficiency and exogenous administration of vasopressin raises blood pressure by increasing vascular tone.¹³ By reducing the requirement for catecholamines, it decreases the stimulation of arrhythmogenic myocardial β₁-receptors and associated myocardial oxygen demand.^7,14 This, among other mechanisms, may translate into a reduction in adverse events, including atrial fibrillation, injury to other organs, and death.^7,15 The most recent Surviving Sepsis guidelines suggest adding vasopressin to norepinephrine to raise MAP to target, or adding vasopressin to decrease norepinephrine dosage (weak recommendations, moderate quality of evidence).⁵

The objective of this systematic review and meta-analysis was to determine the association between treatment with vasopressin in addition to catecholamine vasopressors on atrial fibrillation, morbidity, and mortality compared with catecholamines alone.

Methods

The study protocol was registered with PROSPERO (2017:CRD42017059058). The conduct and reporting of the study follow the PRISMA guidelines.

Eligibility Criteria

Randomized clinical trials were included, irrespective of publication status, date of publication, risk of bias, outcomes published, or language. Trials were included if they enrolled adults with distributive shock, including septic shock, post–cardiovascular surgery vasoplegia, neurogenic shock, and anaphylaxis. Included studies had to compare the administration of vasopressin (or analogues [eg, terlipressin, selepressin]) with or without concomitant catecholaminergic vasopressors with the administration of catecholaminergic vasopressors alone, irrespective of dose, duration, or co-intervention.

The primary outcome was atrial fibrillation. Secondary outcomes were mortality, requirement for renal replacement therapy (RRT), myocardial injury, ventricular arrhythmia, stroke, and LOS in the intensive care unit (ICU) and hospital (eAppendices 4-5 in the Supplement). Acute kidney injury and digital ischemia were post hoc outcomes.

The outcomes were accepted as defined by study authors. For mortality, mortality at 28 to 30 days, at longest follow-up, and in-hospital were considered equivalent; ICU mortality was not pooled. Under digital ischemia, limb ischemia and peripheral ischemia or cyanosis were included. Myocardial infarction, myocardial ischemia, troponin rise, and acute coronary syndrome were pooled under myocardial injury. Ventricular tachycardia and fibrillation were pooled as ventricular arrhythmia. Cerebrovascular accident was combined with stroke.

Search Methods

MEDLINE, EMBASE, and CENTRAL were searched for keywords describing the condition, intervention, or comparator from inception to February 25, 2018 (eAppendices 1-3 in the Supplement). An information specialist reviewed the search strategies.

Trial registries were searched for ongoing and unpublished clinical trials via http://www.isrctn.com using the multiple database search option metaRegister of Controlled Trials and the World Health Organization trial registry. Authors hand-searched the conference proceedings for the scientific sessions of the European Society of Intensive Care Medicine, the Society of Critical Care Medicine, and the American Thoracic Society in the last 2 years. The references of eligible papers were screened and experts were consulted to identify additional trials.

Data Collection and Analysis

Two reviewers independently screened studies’ titles and abstracts for eligibility. Full papers of the potentially eligible studies were retrieved. The same 2 reviewers then independently screened full texts in duplicate and recorded the main reason for exclusion. Disagreements were resolved through discussion.

Data Extraction and Management

Independently, 2 reviewers abstracted data on intervention and outcome. They also recorded study and patient characteristics including age, sex, type of shock, and concomitant conditions (eg, cirrhosis, malignancy). They compared results and resolved disagreements by discussion with a third party. Authors were contacted to clarify ambiguities and to request data on outcomes missing in primary reports.

Assessment of Risk of Bias

In duplicate, 2 review authors assessed risk of bias.¹⁶ In each trial, reviewers evaluated the following domains: sequence generation, allocation concealment, blinding of patients and personnel, blinding of outcome assessors, incomplete outcome data, and selective reporting. The results were compared and disagreements resolved by discussion. Performance and detection bias were assessed separately. All open-label studies were classified as being at high risk of performance bias. A priori, the decision was made to classify open-label designs as “likely low risk of bias” for detection bias for mortality, stroke, and LOS in the absence of other concerns, but to judge “likely high risk of bias” for detection bias for atrial fibrillation, RRT, digital ischemia, myocardial injury, and ventricular arrhythmia. For analysis and presentation purposes, risk of bias was dichotomized as high (or likely high) or low (or likely low).

For subgroup analyses, the study-level risk of bias was assessed for each outcome. If a study was at risk of selection, performance, detection, or reporting bias for that outcome, it was categorized as high risk of bias. Additionally, studies at risk of attrition bias were categorized as high risk of bias for mortality.

Measures of Association With Treatment

The main reported standard association measure for clinical outcomes was risk ratios (RRs) and mean differences for LOS. Risk difference and absolute risk difference were also calculated for clinical outcomes. The absolute risk difference was obtained by applying the RRs with 95% CIs to the baseline risk in the control group. To permit meta-analysis, if a study reporting on LOS provided a median and a measure of dispersion, this was converted to mean and standard deviation assuming a normal distribution.¹⁷

Clinical and methodological heterogeneity were assessed based on study characteristics. Statistical heterogeneity was measured with the I² statistic. An I² statistic greater than 50% was considered as showing substantial heterogeneity.¹⁶

RevMan (Cochrane Collaboration), version 5.3, was used to combine data quantitatively when clinical heterogeneity was nonsubstantial. A random-effects model with Mantel-Haenszel weighting was used because several comparisons were expected to show heterogeneity. After recognizing that a substantial proportion of the weight for atrial fibrillation was contributed by a single study,¹⁸ we combined data for this outcome with a fixed-effect model in a sensitivity analysis. For trials in which patients crossed over to the other treatment, the analysis was according to their first assigned group (intention-to-treat principle). Two-sided P values less than .05 were considered statistically significant.

Subgroup and Sensitivity Analyses

Prespecified subgroup analyses were performed hypothesizing that patients with sepsis would derive greater benefit vs cardiovascular surgery. As a separate sensitivity analysis for RRT, the outcome definition was changed to acute kidney injury, as defined by study authors. P values for interaction between subgroups were tested.

Assessment of the Quality of the Evidence

The GRADE (Grades of Recommendation, Assessment, Development, and Evaluation) approach¹⁹ was used to grade the quality of evidence. GRADE appraises the confidence in estimates of effect by considering within-study risk of bias, directness of the evidence, heterogeneity of the data, precision of effect estimates, and risk of publication bias. Funnel plots of standard errors vs effect estimates were inspected for publication bias and small-study effects.

Results

Screening

The electronic search resulted in 1210 unique citations (Figure 1). After reference and full-text screening, 23 studies met eligibility criteria. Details on excluded and included studies and 3 potentially relevant ongoing studies are available (eAppendices 6-8 in the Supplement).

Figure 1. Flow of Study Selection for Trials Comparing Vasopressin + Catecholamines vs Catecholamines Alone for Patients With Distributive Shock.

Included Studies

The 23 studies that compared vasopressin in combination with catecholamines vs catecholamines alone included 3088 patients (mean age, 61.1 years [14.2]; women, 45.3%) (Table 1). Five trials were multicenter.^{30,33,37,39,40} Twenty-two studies included patients with septic shock.^{20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41} Two studies evaluated patients with post–cardiac surgery vasoplegia.^18,28 Vasopressin was the intervention in 13 trials,^{18,23,24,27,28,29,30,33,34,35,36,37,39} whereas 9 studied terlipressin,^{20,21,22,25,26,32,35,38,41} 1 studied selepressin,⁴⁰ and 1 studied pituitrin (a mixture of vasopressin and oxytocin).³¹ One 3-group study compared vasopressin vs terlipressin vs norepinephrine alone.³⁵ Five studies were published only as abstracts.^{18,21,27,36,38}

Table 1. Characteristics of Included Randomized Clinical Trials Comparing Vasopressin + Catecholamines vs Catecholamines Alone in Patients With Distributive Shock.

Source	Design	Setting	No. of Patients	Condition	Treatment Group(s)	Comparison Group(s)	Planned Follow-up	Risk of Bias for Atrial Fibrillation
Abdullah et al,20 2012	Open label	Single center	34	Paracentesis-induced vasodilatory shock and end-stage liver disease	Terlipressin: 1 mg over 30 min, then continuous infusion of 2 μg/kg/min titrated up and weaned within 24 h	NE: 0.1 μg/kg/min titrated up and weaned within 24 h	48 h	High
Acevedo et al,21 2009	Open label	Single center	24	Septic shock and cirrhosis	Terlipressin: 1-2 mg over 4 h	Adrenergic drugs as needed	Hospital LOS	High
Albanese et al,22 2005	Open label	Single center	20	Septic shock and 2 or more organ dysfunctions	Terlipressin: 1-mg bolus, then another 1-mg bolus if MAP <65 mm Hg	NE: 0.3 μg/kg, then increased by 0.3 μ/kg every 4 min until MAP between 65-75 mm Hg	Hospital LOS	High
Barzegar et al,23 2014	Open label	Single center	30	Septic shock	Vasopressin: 0.03 U/min	NE as needed for MAP >65 mm Hg	28 d	High
Capoletto et al,24 2017	Double-blind	NA	107	Septic shock and cancer	Vasopressin: not described	NE: not described	90 d	Low
Chen et al,25 2017	Open label	Single center	57	Acute respiratory distress syndrome and septic shock	Terlipressin and NE: 0.01-0.04 U/min of terlipressin and NE as needed to maintain MAP between 65-75 mm Hg	NE: >1 μg/min	28 d	High
Choudhury et al,26 2016	Open label	Single center	84	Septic shock and cirrhosis	Terlipressin: 1.3-5.2 μg/min over 24 h	NE: 7.5-60 μg/min	28 d	High
Clem et al,27 2016	Open label	Single center	82	Septic shock	Vasopressin and NE: 0.05-0.5 μg/kg/min of NE and 0.04 U/min of vasopressin to maintain MAP between 65-75 mm Hg	NE: 0.05-0.5 μg/kg/min to maintain MAP between 65-75 mm Hg	28 d	High
Dünser et al,28 2003	Open label	Single center	48	Vasodilatory shock including septic shock, sepsis, and cardiotomy	Vasopressin: 4 U/h at a constant rate	NE: as needed for MAP >70 mm Hg, and additional vasopressin for NE requirements >2.26 μg/kg/min	ICU LOS	High
Fonseca Ruiz et al,29 2013	Open label	Single center	30	Septic shock	Vasopressin and NE: NE + vasopressin at titrated doses of 0.01 U/min, then increased by 0.01 U/min every 10 min to achieve MAP >65 mm Hg or until maximum dose of 0.04 U/min	NE	28 d	High
Gordon et al,30 2016	Double-blind	Multicenter	421	Septic shock	Vasopressin: up to 0.06 U/min with target MAP between 65-75 mm Hg, or at physician’s discretion	NE: up to 12 μg/min with target MAP between 65-75 mm Hg, or at physician’s discretion	28 d	Lowa
Hajjar et al,18 2017	Double-blind	Single center	330	Vasoplegia after cardiac surgery	Vasopressin: 0.01-0.06 U/min with MAP >65 mm Hg	NE: 10-60 μg/min with MAP >65 mm Hg	30 d	Lowa
Han et al,31 2012	Open label	Single center	139	Septic shock	Pituitrin: 1.0-2.5 U/h	NE: 2-20 μg/kg/min	28 d	High
Hua et al,32 2013	Open label	Single center	32	Septic shock and acute respiratory distress syndrome	Terlipressin: 1.3 μg/kg/h	Dopamine: 20 μg/kg/min	28 d	High
Lauzier et al,33 2006	Open label	Multicenter	23	Septic shock	Vasopressin: 0.04-0.20 U/min	NE: 0.1-2.8 μg/kg/min	ICU LOS	High
Malay et al,34 1999	Double-blind	Single center	10	Septic shock	Vasopressin: 0.04 U/min	Placebo	24 h	Low
Morelli et al,35 2009b	Open label	Single center	45	Septic shock	Group 1: vasopressin: 0.03 U/min continuously over 48 h	NE as needed	ICU LOS	High
					Group 2: terlipressin: 1.3 μg/kg/min continuously over 48 h
Oliveira et al,36 2014	Double-blind	Single center	387	Septic shock	Vasopressin: 0.01-0.03 U/min	NE: 0.05-2.0 μg/kg/min	28 d	High
Patel et al,37 2002	Double-blind	Multicenter	24	Septic shock	Vasopressin: 0.01-0.08 U/min	NE: 2-16 μg/min	4 h	Low
Prakash et al,38 2017	Open label	NS	184	Cirrhosis and sepsis	Terlipressin and NE: 2 mg/24 h fixed dose infusion of terlipressin and 3.75-30 μg/min of NE as needed to maintain MAP >65 mm Hg	NE: 7.5-60 μg/min	30 d	High
Russell et al,39 2008	Double-blind	Multicenter	802	Septic shock	Vasopressin: 0.01 U/min, then titrated up to 0.03 U/min with target MAP between 65-75 mm Hg, or at physician’s discretion	NE: 5 μg/min to 15 μg/min with target MAP between 65-75 mm Hg, or at physician’s discretion	90 d	Low
Russell et al,40 2017	Double blind	Multicenter	53	Septic shock	Selepressin: 1.25, 2.5, or 3.75 ng/kg/min until shock resolution or a maximum of 7 d	Placebo	28 d	Lowa
Svoboda et al,41 2012	Open label	Single center	32	Septic shock	Terlipressin: 4 mg over 24 h for 72 h	NE as needed	28 d	High

Abbreviations: ICU, intensive care unit; LOS, length of stay; MAP, mean arterial pressure; NE, norepinephrine or noradrenaline; NS, not specified.

^aTrial judged to be at high risk of bias for mortality due to violation of the intention-to-treat principle.

^bMorelli et al, 2009,³⁵ comprised 3 groups (vasopressin vs terlipressin vs norepinephrine). It was considered as 2 separate trials (vasopressin vs norepinephrine and terlipressin vs norepinephrine) in the comparison between vasopressin and vasopressin analogs. It was considered as a single trial (vasopressin or terlipressin vs norepinephrine) in all other comparisons.

Risk of Bias

Fifteen of 23 trials were not blinded (eAppendices 9-10 in the Supplement). Performance bias due to lack of blinding was judged to have an important effect on all outcomes; patients with distributive shock are critically ill and receiving many concomitant interventions that could be influenced by choice of concomitant vasopressor. Atrial fibrillation, myocardial injury, and digital ischemia are vulnerable to detection bias from differential capture and subjective interpretation; lack of blinding of clinicians and outcome assessors may influence these outcomes. The decision to start RRT could also be subjective. Other outcomes were judged to be at low risk of detection bias in the absence of blinding. Two studies were assessed to be at risk of selection bias due to inadequate randomization^31,36; they did not describe their randomization process and had significant between-group imbalances. Nine studies (39%) reported the information necessary to make a definitive judgment for selection bias. Authors relied on imbalances between groups and overall methodological quality of the study to make this judgment. Attrition was found in 7 studies,^{18,25,30,31,36,40,41} and judged as having an effect on mortality (Table 2 and Figure 2A). Reporting bias was not detected. “Other bias” was judged to be present when studies were published as abstracts only. Prespecified sensitivity analyses were performed to assess the robustness of estimates to risk of bias if studies were dichotomized according to their risk of bias.

Table 2. Association of Vasopressin + Catecholamine Vasopressors vs Catecholamines Alone With Atrial Fibrillation in Patients With Distributive Shock and Sensitivity Analyses.

Group	No. With Events/Total No. of Patients		Risk Difference, % (95% CI)a	Relative Riska			Quality of Evidence
	Vasopressin + Catecholamines	Catecholamines Alone		Risk Ratio (95% CI)	P Value	I2 %
All studies18,20,24,26,27,28,30,33,34,35,39,40,41	159/739	215/723	−6 (−13 to 1)	0.77 (0.67 to 0.88)	<.001	1	High
Low risk of bias18,24,30,34,39,40	136/559	182/554	−7 (−20 to 5)	0.77 (0.68 to 0.88)	<.001	0
High risk of bias20,26,27,28,33,35,41	23/180	33/169	−3 (−10 to 4)	0.73 (0.40 to 1.34)	.31	36
Sepsis20,24,26,27,28,30,33,34,35,39,40,41,b	60/580	84/563	−3 (−7 to 1)	0.76 (0.55 to 1.05)	.09	8
Cardiac surgery18,28,c	99/159	131/160	−19 (−29 to −10)	0.77 (0.67 to 0.88)	<.001	0
Vasopressin18,24,27,28,30,33,34,35,39,b,c	151/621	201/626	−7 (−17 to 3)	0.77 (0.68 to 0.88)	<.001	0
Vasopressin analogues20,26,35,40,41,c	8/118	18/112	−0.05 (−11 to 1)	0.52 (0.18 to 1.51)	.23	28
Fixed-effect analysis18,20,24,26,27,28,30,33,34,35,39,40,41,b,c	159/739	215/723	−7 (−11 to −4)	0.75 (0.65 to 0.86)	<.001	1

^aRelative risk <1.0 and risk difference <0.0 favors vasopressin + catecholamines.

^bDünser et al, 2003,²⁸ included patients with both sepsis and post–cardiac surgery vasoplegia, but subgroup data were obtained for atrial fibrillation only. This study was excluded from other outcomes when sepsis and post–cardiac surgery vasoplegia were compared.

^cMorelli et al, 2009,³⁵ comprised 3 groups (vasopressin vs terlipressin vs norepinephrine). It was considered as 2 separate trials (vasopressin vs norepinephrine and terlipressin vs norepinephrine) in the comparison between vasopressin and vasopressin analogs. It was considered as a single trial (vasopressin or terlipressin vs norepinephrine) in all other comparisons.

Figure 2. Relative Risks of All Trials Comparing Vasopressin + Catecholamines vs Catecholamines Alone for Patients With Distributive Shock.

The relative risks were calculated using a random-effects model with Mantel-Haenszel weighting. The size of data markers indicates the weight of the study. Error bars indicate 95% CIs.

^aVasopressin (or analogue [ie, terlipressin, selepressin, or pituitrin]) + catecholamine vasopressors.

Primary Outcome

Atrial Fibrillation

Pooling data from 13 studies (4 studies with 0 events in either group, 1462 patients, 374 events) demonstrated a significant reduction in the risk of atrial fibrillation associated with the administration of vasopressin (RR, 0.77 [95% CI, 0.67 to 0.88], I² = 1%; risk difference [RD], −0.06 [95% CI, −0.13 to 0.01]) (Figure 2A). Based on the GRADE framework, this was judged to be high-quality evidence (eAppendix 13 in the Supplement). This result was driven by the study by Hajjar et al,¹⁸ which carried 74.8% of the weight. In absolute terms, the absolute effect is that 68 fewer people per 1000 patients (95% CI, 36 to 98) will experience atrial fibrillation when vasopressin is added to catecholaminergic vasopressors. In a sensitivity analysis excluding the 7 studies at high risk of bias for lack of blinding of outcome assessors,^{20,26,27,28,33,35} the estimate of effect was unchanged (eAppendix 11 in the Supplement). In a second sensitivity analysis, patients with sepsis and post–cardiac surgery were considered separately. For the subgroup of post–cardiac surgery patients,^18,28 the resultant RR estimate was 0.77 (95% CI, 0.67 to 0.88), not significantly different than in patients with sepsis (RR, 0.76 [95% CI, 0.55 to 1.05], P for interaction = .97) (Table 2). Even though the crude rate of atrial fibrillation in this post–cardiac surgery population (73%) was considerably higher than in the sepsis studies (13%), the relative effect estimate was similar in both groups.

Secondary Outcomes

Mortality

Mortality data were available from 17 studies (2904 patients, 1123 events) (Figure 2B). When pooled, the administration of vasopressin in addition to catecholamines was associated with a reduction in mortality (RR, 0.89 [95% CI, 0.82 to 0.97], P = .009, I² = 0; RD, −0.04 [95% CI, −0.07 to 0.00]). In absolute terms, 45 lives (95% CI, 12 to 73) would be saved per 1000 patients treated with vasopressin. However, when limited to the 2 trials at low risk of bias (Table 3),^24,39 the RR estimate was 0.96 (95% CI, 0.84 to 1.11).

Table 3. Binary Outcomes and Sensitivity Analyses for Vasopressin + Catecholamines vs Catecholamines Alone in Patients With Distributive Shock.

Group	No. With Events/Total No. of Patients		Risk Difference % (95% CI)a	Relative Riska			Quality of Evidence (Reason for Judgment)
	Vasopressin + Catecholamines	Catecholamines Alone		Risk Ratio (95% CI)	P Value	I2 %
28-d or 30-d Mortality
All studies18,21,22,23,24,25,26,27,29,30,31,32,36,38,39,40,41	532/1453	591/1451	−4 (−7 to 0)	0.89 (0.82 to 0.97)	.009	0	Low (risk of bias)
Low risk of bias24,39	215/529	222/520	−2 (−8 to 4)	0.96 (0.84 to 1.11)	.6	0
High risk of bias18,21,22,23,25,26,27,29,30,31,32,36,38,40,41	317/924	369/931	−4 (−8 to 0)	0.86 (0.77 to 0.95)	.004	0
28-d or 30-d or ICU mortality18,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,38,39,40,41,b,c	567/1525	623/1505	−4 (−7 to −1)	0.89 (0.83 to 0.97)	.006	0
Full text only18,22,23,25,26,29,30,31,32,39,40,41,d	334/993	356/984	−2 (−6 to 2)	0.91 (0.82 to 1.01)	.09	0
Vasopressin23,24,27,29,30,36,39,41,b	404/1156	431/1160	−2 (−6 to 2)	0.94 (0.85 to 1.04)	.21	0
Vasopressin analogues21,22,25,26,31,32,38,40,41,b	128/297	160/291	−10 (−18 to −3)	0.81 (0.70 to 0.94)	.005	0
Sepsis21,22,23,24,25,26,27,29,30,31,32,36,38,39,40,41	509/1304	567/1300	−4 (−8 to −1)	0.89 (0.82 to 0.97)	.008	0
Cardiac surgery18	23/149	24/151	−0 (−9 to 8)	0.97 (0.57 to 1.64)	.91	NA
Requirement for Renal Replacement Therapy
All studies23,24,28,30,33,35,b,e	97/412	125/393	−7 (−12 to −1)	0.74 (0.51 to 1.08)	.12	70	Moderate (imprecision)
Low risk of bias24,30	62/330	89/329	−7 (−13 to −2)	0.70 (0.53 to 0.92)	.01	0
High risk of bias23,28,33,35,b,c	35/82	36/64	−5 (−16 to 7)	0.77 (0.42 to 1.43)	.41	67
AKI as outcome18,21,24,28,30,b	154/515	204/516	−8 (−21 to 6)	0.73 (0.46 to 1.17)	.19	91
Vasopressin23,24,28,30,33,35,b,e	93/397	125/393	−6 (−11 to −1)	0.76 (0.53 to 1.10)	.15	68
Vasopressin analogues35,b,e	4/15	8/15	−27 (−60 to 7)	0.50 (0.19 to 1.31)	.16	NA
Digital Ischemia
All studies18,23,24,26,29,30,39,40,41	41/990	17/973	2 (−1 to 4)	2.38 (1.37 to 4.12)	.002	0	Moderate (post hoc outcome)
Low risk of bias18,24,30,39,40	23/906	9/883	1 (−1 to 3)	2.45 (1.10 to 5.43)	.03	0
High risk of bias23,26,29,41	18/84	8/90	10 (0 to 19)	2.31 (1.08 to 4.94)	.03	0
Defined as digital ischemia18,23,29,30,33,39,40,f	25/810	8/789	2 (0 to 3)	2.73 (1.27 to 5.87)	.01	0
Vasopressin18,23,24,29,30,33,39,b	24/904	10/893	1 (−1 to 3)	2.35 (1.10 to 5.05)	.03	0
Vasopressin analogues26,40,41,b	17/86	7/80	10 (−4 to 25)	2.40 (1.09 to 5.31)	.03	0
Myocardial Injury
All studies18,20,24,28,30,33,34,37,39,40,41,b	62/991	71/966	0 (−2 to 2)	0.86 (0.63 to 1.17)	.34	0	Low (indirectness, imprecision)
Low risk of bias18,24,30,34,37,39,40	61/924	66/899	1 (−1 to 3)	0.89 (0.64 to 1.25)	.52	4
High risk of bias20,28,33,41,b	1/67	5/67	−5 (−12 to 3)	0.37 (0.07 to 1.95)	.24	0
Sepsis20,24,28,30,33,34,37,39,40,41,b	51/818	51/791	1 (−1 to 2)	0.94 (0.67 to 1.32)	.71	0
Cardiac surgery18	11/149	17/151	−4 (−10 to 3)	0.66 (0.32 to 1.35)	.25	NA
Vasopressin18,24,28,30,33,34,37,39,b	61/930	70/912	0 (−3 to 2)	0.87 (0.61 to 1.23)	.42	6
Vasopressin analogues20,40,41,b	1/61	1/54	1 (−6 to 7)	0.91 (0.10 to 8.33)	.93	0
Ventricular Arrhythmia
All studies18,20,24,26,27,33,34,37,41	39/418	48/419	0 (−2 to 1)	0.93 (0.73 to 1.19)	.55	0	Low (indirectness, imprecision)
Low risk of bias18,34,37	27/167	32/167	−2 (−10 to 5)	0.86 (0.54 to 1.35)	.50	NA
High risk of bias20,24,26,27,33,41	12/251	16/252	0 (−1 to 1)	0.96 (0.72 to 1.28)	.78	0
Vasopressin18,24,27,33,34,37,b	28/346	32/343	0 (−1 to 2)	0.88 (0.56 to 1.38)	.57	0
Vasopressin analogues20,26,41,b	11/72	16/76	−2 (−7 to 3)	0.95 (0.71 to 1.27)	.73	0
Stroke
All studies18,24,39,41	11/683	6/675	1 (−2 to 4)	1.61 (0.53 to 4.95)	.40	7	Moderate (imprecision)
Low risk of bias18,24,39	11/670	6/658	1 (−2 to 4)	1.61 (0.53 to 4.95)	.40	7
High risk of bias41	0/13	0/17	0 (−12 to 12)	NA	NA	NA
Vasopressin18,24,39,b	11/670	6/658	1 (−2 to 4)	1.61 (0.53 to 4.95)	.40	7
Vasopressin analogues41,b	0/13	0/17	0 (−12 to 12)	NA	NA	NA

Abbreviation: AKI, acute kidney injury.

^aRelative risk <1.0 and risk difference <0.0 favors vasopressin + catecholamines.

^bDünser et al, 2003,²⁸ included patients with both sepsis and post–cardiac surgery vasoplegia, but subgroup data were obtained for atrial fibrillation only. This study was excluded from other outcomes when sepsis and post–cardiac surgery vasoplegia were compared.

^cAdded 4 studies that reported on ICU mortality.

^d“Full text only” refers to studies not published only as abstracts.

^eMorelli et al, 2009,³⁵ comprised 3 groups (vasopressin vs terlipressin vs norepinephrine). It was considered as 2 separate trials (vasopressin vs norepinephrine and terlipressin vs norepinephrine) in the comparison between vasopressin and vasopressin analogs. It was considered as a single trial (vasopressin or terlipressin vs norepinephrine) in all other comparisons.

^fIncludes only studies in which the authors described the outcome as digital ischemia. Peripheral cyanosis and limb ischemia were excluded.

Requirement for RRT and Acute Kidney Injury

Six trials with a total of 805 patients (222 events) reported on RRT (Figure 3A). When combined, vasopressin was associated with a reduced risk for RRT, but the pooled estimate did not reach statistical significance and showed substantial heterogeneity (RR, 0.74 [95% CI, 0.51 to 1.08], I² = 70%; RD, −0.07 [95% CI, −0.12 to −0.01]; moderate-quality evidence). However, when the analysis was limited to the 2 trials at low risk of bias,^24,30 the point estimate was similar, but vasopressin was associated with a significant reduction in the risk of RRT without evidence of heterogeneity (RR, 0.70 [95% CI, 0.53 to 0.92], I² = 0%, P for interaction = .77).

Figure 3. Relative Risks of All Trials Comparing Vasopressin + Catecholamines vs Catecholamines Alone for Patients With Distributive Shock.

The relative risks were calculated using a random-effects model with Mantel-Haenszel weighting. The size of data markers indicates the weight of the study. Error bars indicate 95% CIs.

^aVasopressin (or analogue [ie, terlipressin, selepressin, or pituitrin]) + catecholamine vasopressors.

^bRisk of bias categories for requirement for renal replacement therapy are the same as those for atrial fibrillation, as summarized in Table 1.

Myocardial Injury

Eleven studies (1957 patients, 133 events) reported on myocardial injury; 2 trials had event rates of 0 in both groups. There was no significant difference in the risk of myocardial injury with vasopressin (RR, 0.86 [95% CI, 0.63 to 1.17], I² = 0%; RD, 0.00 [95% CI, −0.02 to 0.02]; low-quality evidence). After excluding studies at high risk of bias from open-label design,^20,28,33,41 the estimate did not change significantly. Because surrogates were reported for myocardial injury (eg, altered ST segments), indirectness was rated as serious.

Ventricular Arrhythmia, Stroke, and Length of Stay

When 9 studies reporting on ventricular arrhythmia were pooled (837 patients, 87 events), the risk was not significantly different with vasopressin (RR, 0.93 [95% CI, 0.73 to 1.19], I² = 0%; RD, 0.00 [95% CI, −0.02 to 0.01]; low-quality evidence). There was no significant difference when pooling data from 4 studies (1358 patients, 17 events) reporting on stroke (RR, 1.61 [95% CI, 0.53 to 4.95], I² = 7%; RD, 0.01 [95% CI, −0.02 to 0.04]; moderate-quality evidence). LOS data were reported exclusively as medians with interquartile range and were transformed to estimate mean LOS with standard deviation. Hospital LOS was not significantly associated with vasopressin (8 studies, 1939 patients; mean difference, −1.14 days [95% CI, −3.60 to 1.32], I²=75%; low-quality evidence) (Table 4). Similarly, ICU LOS was not significantly associated with vasopressin (mean difference, −0.40 days [95% CI, −1.05 to 0.25], I²= 24%; moderate-quality evidence) when 11 studies were combined (2156 patients).

Table 4. Continuous Outcomes and Sensitivity Analyses for Vasopressin + Catecholamines vs Catecholamines Alone in Patients With Distributive Shock.

Group	Mean Length of Stay in Days (SD)a		Mean Difference (95% CI), db	P Value	I2 %	Quality of Evidence (Reason for Judgment)
	Vasopressin + Catecholamines	Catecholamines Alone
Hospital Length of Stay
All studies18,20,22,29,32,34,38,40	21.3 (23.0)	22.6 (22.9)	−1.14 (−3.60 to 1.32)	.36	75	Low (imprecision, inconsistency)
Low risk of bias18,24,30,39	22.0 (24.1)	23.3 (23.8)	−1.83 (−4.47 to 0.81)	.17	69
High risk of bias29,32,34,40	15.7 (8.6)	16.6 (11.1)	−0.45 (−4.40 to 3.50)	.82	62
Vasopressin18,24,29,30,39,c	21.8 (24.0)	23.4 (23.7)	−2.33 (−5.05 to 0.40)	.09	67
Vasopressin analogs29,32,40,c	16.1 (7.7)	14.9 (8.5)	1.03 (−1.48 to 3.53)	.42	22
Intensive Care Unit Length of Stay
All studies18,20,22,28,29,31,32,35,38,39	11.1 (12.2)	11.6 (13.4)	−0.40 (−1.05 to 0.25)	.23	24	Moderate (imprecision)
Low risk of bias18,24,30,39	11.2 (12.7)	12.2 (14.1)	−0.54 (−1.33 to 0.25)	.18	34
High risk of bias28,29,31,32,35,39,40	10.4 (10.2)	9.4 (8.5)	−0.12 (−1.37 to 1.13)	.85	22
Vasopressin18,23,24,28,30,35,39,c	11.8 (13.0)	12.4 (14.0)	−0.24 (−1.27 to 0.79)	.65	44
Vasopressin analogues29,31,32,35,40,c	7.9 (7.0)	7.9 (7.0)	−0.38 (−1.33 to 0.58)	.44	0

^aMean length of stay was weighted by the number of patients.

^bMean difference <0.0 favors vasopressin + catecholamines.

^cMorelli et al, 2009,³⁵ comprised 3 groups (vasopressin vs terlipressin vs norepinephrine). It was considered as 2 separate trials (vasopressin vs norepinephrine and terlipressin vs norepinephrine) in the comparison between vasopressin and vasopressin analogs. It was considered as a single trial (vasopressin or terlipressin vs norepinephrine) in all other comparisons.

Post Hoc Outcomes

Digital Ischemia

When 9 studies (1963 patients, 58 events) were pooled (Figure 3B), vasopressin in addition to catecholamines was associated with a significant increase in digital ischemia (RR, 2.38 [95% CI, 1.37 to 4.12], I² = 0; RD, 0.02 [95% CI, −0.01 to 0.04]; moderate-quality evidence). In absolute terms, this means 24 more occurrences (95% CI, 6 to 55) of digital ischemia per 1000 patients treated with vasopressin. When the 4 studies at high risk of bias were excluded,^23,26,29,41 the resultant estimate was not significantly different. Definitions varied for this outcome; however, when the analysis was limited to the 6 studies that specifically described “digital ischemia,”^{18,23,29,30,39,40} the resultant estimate did not change significantly. Thus, evidence was not downgraded for indirectness but, because it was a post hoc outcome, it was downgraded for risk of bias.

Acute Kidney Injury

In a sensitivity analysis, the treatment effect for RRT was consistent, but not statistically significant when the definition was modified to acute kidney injury (5 trials; RR, 0.73 [95% CI, 0.46 to 1.17], I² = 91%).

Publication Bias

The assessment of publication bias was limited by small numbers of studies for most outcomes (eAppendix 12 in the Supplement). Visual inspection did not lead to concerns about publication bias.

Discussion

In this systematic review and meta-analysis of randomized clinical trials, the administration of vasopressin in addition to catecholamine vasopressors in patients with distributive shock was associated with a significant reduction in the risk of atrial fibrillation when compared with catecholamines alone (high-quality evidence). Findings for other outcomes were not consistent. Although when all studies were combined the risk of mortality was lower with the addition of vasopressin, a sensitivity analysis limited to low risk of bias trials yielded a relative risk much closer to 1 and was not statistically significant.

To our knowledge, this systematic review is the first on the topic to include atrial fibrillation as an outcome. Prior reviews assessed arrhythmia,^42,43 but this outcome has limited utility due to the variety of conditions that could be found under this heading. The reduction in atrial fibrillation associated with vasopressin was consistent across 2 subtypes of distributive shock and in sensitivity analyses restricted to studies at low risk of bias.

Vasopressin may have contributed to a reduction of atrial fibrillation by sparing the adrenergic stimulation provided by catecholaminergic vasopressors.^6,7,8,14 This could have manifested in fewer patients developing atrial fibrillation or may have caused atrial fibrillation to be shorter in duration and lower in rate and, in consequence, less likely to be detected.

The approach to monitoring and ascertainment of atrial fibrillation in patients who are acutely ill affects the detection of this outcome.⁴⁴ This limitation would need to be addressed to more precisely estimate event rates in this population and their association with vasopressin treatment. The clinical significance of atrial fibrillation in this population is not fully understood.⁴⁴ Where atrial fibrillation in patients who are critically ill has been associated with worse outcomes, including death, causality has not been proven and the consequences on long-term prognosis in survivors are unknown.^10,11,44

This review is one of few reviews to directly compare vasopressin + catecholamines against the current standard of care—catecholamines alone. Two systematic reviews with network meta-analyses found no difference in mortality in any comparison, including between vasopressin or terlipressin and norepinephrine.^42,43 Another systematic review and meta-analysis concluded that treatment with noncatecholaminergic agents (including vasopressin and methylene blue) improved survival (RR, 0.88 [95% CI, 0.79 to 0.98]) in patients experiencing or “at risk” for distributive shock.⁴⁵ In another systematic review and meta-analysis, mortality was significantly lower in patients with septic shock treated with vasopressin or terlipressin compared with norepinephrine (RR, 0.87 [95% CI, 0.78 to 0.97]).⁴⁶ However, that review included 4 substudies of the Vasopressin and Septic Shock Trial (VASST) in the meta-analysis of mortality and did not assess evidence using GRADE.^19,39,47

The theoretical basis for vasopressin administration stems from research identifying relative vasopressin deficiency in patients with distributive shock.¹³ Vasopressin administration could lower mortality by decreasing the need for catecholaminergic drugs and reducing their adverse effects including arrhythmia, preferentially perfusing the brain and renal vascular bed—the latter leading to reductions in acute kidney injury—and decreasing activation of both the renin-aldosterone-angiotensin system and neurohormonal processes, inhibiting proinflammatory cytokines, improving calcium handling, and potentiating endogenous glucocorticoids.^47,48,49,50

For clinicians aiming for MAP, maintaining adequate blood flow while mitigating the risk of excessive vasoconstriction (the likely mechanism of digital ischemia) is also important. An understanding of the clinical effect of these events (ie, did they simply precipitate drug discontinuation or did they lead to permanent disability?) would be needed to evaluate trade-off against a decrease in mortality.

This systematic review also evaluated requirement for RRT. The significant reduction in need for RRT with vasopressin was limited to the pooled estimate for low risk of bias studies. Renal protection related to reduced activation of the renin-aldosterone-angiotensin system is one of the hypothesized benefits of vasopressin in distributive shock; creatinine clearance has been shown to improve when vasopressin was started early after the onset of distributive shock.³³

This review included data from the relatively large and recently published Vasopressin vs Norepinephrine in Patients with Vasoplegic Shock after Cardiac Surgery (VANCS) and Effect of Early Vasopressin vs Norepinephrine on Kidney Failure in Patients With Septic Shock (VANISH) trials (751 patients total).^18,30 Combining subtypes of distributive shock and considering vasopressin analogs allowed the inclusion of a larger number of studies. Bias in the review process was reduced by searching multiple databases without language restriction. Significant attempts were made to obtain clarification of published data and access to unpublished data.

Limitations

This study has several limitations. First, subgroup analyses were restricted by the study-level nature of the data. Second, the quality of reporting for many studies was not sufficient to permit definitive judgments about risk of bias in all domains. Third, there are likely differences in the way vasopressors were initiated, titrated, and weaned between studies and approaches were infrequently described in detail. However, the general approach seemed to be to up-titrate vasopressin until the maximum dose or target MAP was reached and then to add or wean norepinephrine as needed to reach the target MAP.

Conclusions

In this meta-analysis, the addition of vasopressin to catecholamine vasopressors compared with catecholamines alone was associated with a lower risk of atrial fibrillation. However, findings for secondary outcomes varied.

Supplement.

eAppendix 1. MEDLINE Search Strategy

eAppendix 2. EMBASE Search Strategy

eAppendix 3. Cochrane CENTRAL Search Strategy

eAppendix 4. Basis for Outcome Selection

eAppendix 5. Outcome Importance for Choice of Vasopressor in Patients With Vasodilatory Shock

eAppendix 6. Characteristics of Included Studies

eAppendix 7. Characteristics of Important Excluded Studies

eAppendix 8. Characteristics of Ongoing Studies

eAppendix 9. Risk of Bias Graphs: Review Authors’ Judgments About Each Risk of Bias Item Presented as Percentages Across All 23 Randomized Trials

eAppendix 10. Risk of Bias Summary: Review Authors’ Judgments About Each Risk of Bias Item for Each Included Study

eAppendix 11. Forest Plots for All Outcomes, Including Sensitivity Analyses

eAppendix 12. Funnel Plots

eAppendix 13. Reported lengths of stay in primary studies and transformation of median and interquartile range to mean and standarddeviation

eAppendix 14. Summary of Findings Table

eReferences.