Vasopressors in septic shock: a systematic review and network meta-analysis.

OBJECTIVE: Vasopressor agents are often prescribed in septic shock. However, their effects remain controversial. We conducted a systematic review and Bayesian network meta-analysis to compare the effects among different types of vasopressor agents. DATA SOURCES: We searched for relevant studies in PubMed, Embase, and the Cochrane Library databases from database inception until December 2014. STUDY SELECTION: Randomized controlled trials in adults with septic shock that evaluated different vasopressor agents were selected. DATA EXTRACTION: Two authors independently selected studies and extracted data on study characteristics, methods, and outcomes. DATA SYNTHESIS: Twenty-one trials (n=3,819) met inclusion criteria, which compared eleven vasopressor agents or vasopressor combinations (norepinephrine [NE], dopamine [DA], vasopressin [VP], epinephrine [EN], terlipressin [TP], phenylephrine [PE], TP+NE, TP + dobutamine [DB], NE+DB, NE+EN, and NE + dopexamine [DX]). Except for the superiority of NE over DA, the mortality of patients treated with any vasopressor agent or vasopressor combination was not significantly different. Compared to DA, NE was found to be associated with decreased cardiac adverse events, heart rate (standardized mean difference [SMD]: -2.10; 95% confidence interval [CI]: -3.95, -0.25; P=0.03), and cardiac index (SMD: -0.73; 95% CI: -1.14, -0.03; P=0.004) and increased systemic vascular resistance index (SVRI) (SMD: 1.03; 95% CI: 0.61, 1.45; P<0.0001). This Bayesian meta-analysis revealed a possible rank of probability of mortality among the eleven vasopressor agents or vasopressor combinations; from lowest to highest, they are NE+DB, EN, TP, NE+EN, TP+NE, VP, TP+DB, NE, PE, NE+DX, and DA.
CONCLUSION: In terms of survival, NE may be superior to DA. Otherwise, there is insufficient evidence to suggest that any other vasopressor agent or vasopressor combination is superior to another. When compared to DA, NE is associated with decreased heart rate, cardiac index, and cardiovascular adverse events, as well as increased SVRI. The effects of vasopressor agents or vasopressor combinations on mortality in patients with septic shock require further investigation.

Introduction

Septic shock is a life-threatening condition and severe sepsis accounts for 20% of all admissions to intensive care units.1 Severe sepsis approximates 750,000 cases annually in the USA and has a mortality rate averaging 28%.2 For initial resuscitation, intravenous fluids are recommended as the first-line therapy. However, vasopressor agents are also critical to achieve and maintain adequate blood pressure and tissue perfusion, and hence, should be used early.3 Sakr et al4 reported that the most frequently used vasopressor agent during septic shock was norepinephrine (NE, 80.2%), followed by dopamine (DA, 35.4%), and epinephrine (EN, 23.3%) alone or in combination. Although NE is recommended as the fist-line agent for treating hypotension in volume-resuscitated hyperdynamic septic shock,5 the second-line vasopressor remains controversial. Previous studies have reported that NE may have significant superiority over DA in terms of survival.5–8 However, compared with other vasopressors, such as EN, vasopressin (VP), terlipressin (TP), and phenylephrine (PE), the outcomes on the use of NE were not different. Morelli et al9 reported that there was no difference in terms of cardiopulmonary performance, global oxygen transport, and regional hemodynamics when PE was administered instead of NE in the initial hemodynamic support of septic shock. Russell et al10 revealed that low-dose VP did not improve survival rates in contrast with NE in septic shock patients treated with catecholamine vasopressors. Additionally, EN was recommended as an additional agent to NE to maintain adequate blood pressure.5 Recently in a single-center randomized controlled trial (RCT), NE supplemented by dobutamine (DB) was compared to NE supplemented by EN in the treatment of septic shock patients.11 However, the effectiveness of other vasopressor agents or vasopressor combinations as compared to others is limited. Whether the use of any vasopressor agents or vasopressor combinations in patients with septic shock translates to a survival advantage remains unclear. Meta-analyses of vasopressor agents have been limited by considering only two or three categories of vasopressor agents, not including indirect and direct comparisons, and omission of recent RCTs. Therefore, we performed a network meta-analysis (NMA) considering direct and indirect comparisons of vasopressor agents and vasopressor combinations in reducing overall mortality for septic shock patients.

Materials and methods

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA Statement) guidelines were used to perform this meta-analysis.12

Information sources and eligibility criteria

A search of the PubMed (US National Library of Medicine, Bethesda, MD, USA) and Cochrane Library databases and Embase from database inception to December 2014 was performed. The eligibility criteria were as follows: the study design must be randomized controlled, the study must report mortality outcome, and the study must evaluate adult patients at least 18 years of age.

Search strategy

We used text words and medical subject heading (MeSH) terms with Boolean strategy. The cross-searching was done based on the following three categories: 1) vasopressors related (“vasopressor” or “vasoactive drug” or “catecholamine” or “pressor agent”); 2) different vasopressors (“norepinephrine” or “dopamine” or “epinephrine” or “adrenaline” or “isuprel” or “aleudrin” or “vasopressin” or “terlipressin” or “phenylephrine” or “dopexamine”); 3) disease (“sepsis” or “infection” or “septic shock” or “shock” or “systemic inflammatory response syndrome” or “SIRS”). The search was limited to the “English” language and “human” subjects. Further search by reviewing conference proceedings and the references of review articles was performed manually if necessary.

Study selection

Two independent investigators (FZ and ZM) performed the study selection. Differences between the two investigators were resolved by consensus or adjudicated by a third investigator (XZ). Agreement between the two reviewers on study inclusion was excellent (k=1). Studies on adult patients with septic shock that evaluated the mortality rates of different vasopressor agents or vasopressor combinations were selected.

Data extraction

Two investigators independently extracted raw data using a standard form for each study. The form included year of publication, the study type, number of patients, patient characteristics, and details of the outcomes. The main outcome was 28-day mortality. We used the mortality rate from the only undetermined time point or the nearest time point when mortality was reported at only an undetermined time point or several time points, respectively. In addition, we also assessed cardiac adverse events and hemodynamic and metabolic parameters.

Quality assessment

We assessed the quality of each study selected for this meta-analysis by using the Jadad score, which includes the following criteria: randomization, concealment of treatment allocation, clinician blinding, baseline balance between groups, and the description of withdrawals and dropouts.13

Statistical analysis

A meta-analysis was performed to calculate direct estimates of treatment effect for each pair of vasopressor agents or vasopressor combinations. According to heterogeneity of treatment effect across trials using the I2-statistics,14 a fixed-effect model (P≥0.1) or random-effects model (P<0.1) was used. Results in terms of odds ratio (OR) for dichotomous outcomes or standardized mean difference (SMD) for continuous data were expressed with mean and 95% confidence intervals (CIs). The direct meta-analysis was done using Review Manager, version 5.1.2 (RevMan; The Cochrane Collaboration, Oxford, UK).

Using a Bayesian framework, we performed random-effects NMAs for each vasopressor agent or vasopressor combination. NMA is a recent emerging approach used to evaluate the effect size of all possible pairwise comparisons even if they are not compared head-to-head.15 Results such as ORs are expressed with 95% CIs. These CIs from NMAs are the Bayesian analogs of the 95% CIs.15 The models had 80,000 iterations, while a burn-in of 40,000 and a thin of 10 were used.16 Vague priors were used.16 All convergence on the basis of Brooks–Gelman–Rubin plots was assessed.16 Cumulative probability plot (cumulative probability vs rank curve) is presented. Using R-project 3.1.1, the Z-test was conducted to assess for inconsistency of triangular loops.17 Area under the cumulative probability curve represents the rank of probability. The analysis for the NMA was performed using WinBUGS1.4.3 (Medical Research Council Biostatistics Unit; www.mrc-bsu.cam.ac.uk/software/bugs/) and R-project 3.1.1 (http://cran.r-project.org/). Publication bias was tested by funnel plots whenever possible.

Results

There were 4,280 potentially relevant studies, and 49 articles were retrieved for detailed assessment. Twenty-eight articles were excluded because there were no mortality comparisons (n=20), no sepsis patients (n=2), other septic shock investigations (n=3), and post hoc analyses (n=3). Twenty-one studies were included in this meta-analysis (Figure 1).9–11,18–35 To evaluate hemodynamic outcomes, we extracted heart rate (HR), mean arterial pressure (MAP), systemic vascular resistance index (SVRI), cardiac index, and mortality data from studies by Russell et al10 and Gordon et al.30

Figure 1

Quorum chart of study cohort.

Note: The search had been conducted using the PubMed, Embase, and the Cochrane Library databases from database inception to December 2014.

Study characteristics

Fourteen single-center9,11,18–24,26,29,31,32,34 and seven multicenter studies10,25,27,28,30,33,35 were identified. The characteristics and inclusion criteria of the selected RCTs are summarized in Table 1. These articles were reported between 1993 and 2012, and a total of 3,819 patients were included in this study. Inclusion criteria were not the same for all trials; however, all patients met the diagnosis of severe sepsis or septic shock (Table 1).36 Mean age ranged from 18 years to 70 years, and the proportion of male patients ranged from 46% to 77.3%. The mean Acute Physiology and Chronic Health Evaluation II (APACHE II) score was 23.8.

Table 1

Characteristics of the included randomized trials

Source	Number of patients	Mean age (years)	Male (%)	Center	Mean APACHE II/SAPS II/SOFA score	Blood pressure (mmHg)	Inclusion criteria
Mahmoud and Ammar11	60	51.4	31 (51.7)	S	NR/NR/14.8	MAP <70	Sepsis plus hypotension refractory to an initial fluid challenge
Gordon et al30	241	62.0	146 (60.6)	M	28.2/NR/NR	MAP =72.4	Septic shock with two or more criteria of the systemic inflammatory response syndrome, infection, dysfunction of one or more organs
De Backer et al35	1,044	67.5	NR	M	NR/NR/NR	MAP <70	MAP <70 mmHg or the SBP <100 mmHg after adequate amount of fluids used and signs of tissue hypoperfusion
Patel et al34	252	≥18	116 (46.0)	S	27.5/NR/12	MAP <60 and/or SBP <90	Septic shock requiring vasopressors after adequate fluid used (clinical examination and/or CVP >8 mmHg)
Gordon et al33	778	61.8	475 (61.0)	M	27.1/NR/NR	MAP =72.7 (NE maintaining)	Septic shock with two or more criteria for the systemic inflammatory response syndrome
Jain and Singh32	54	44.1	28 (51.9)	S	18.4/NR/NR	SBP =74.1	SBP <90 mmHg or MAP <60 mmHg and CVP >12 mmHg or PAOP >18 mmHg despite adequate fluid used and continuous dopamine for 1 hour; evidence of one or more end-organ dysfunction, infection with special criteria
Morelli et al31	45	65.7	33 (73.3)	S	NR/60/NR	MAP <65	MAP <65 mmHg despite appropriate volume resuscitation
Morelli et al9	32	70	21 (65.5)	S	NR/56/NR	MAP <65	MAP <65 mmHg despite appropriate volume resuscitation
Morelli et al29	59	66.3	43 (72.3)	S	NR/60/NR	MAP =70 (NE maintaining)	MAP <65 mmHg despite appropriate volume resuscitation
Myburgh et al28	158	≥18	NR	M	NR/NR/NR	MAP ≥60	Clinician judged patients to require either epinephrine or norepinephrine
Russell et al10	778	60.6	475 (61.1)	M	27.1/NR/NR	MAP =72.5 (vaso pressors maintaining)	Septic shock with two or more criteria of the systemic inflammatory response syndrome, infection, one or more organ dysfunction
Annane et al27	330	62.5	202 (61)	M	NR/53/11	MAP <60 and/or SBP <90	Two or more of the systemic inflammatory response syndrome, organ dysfunction, or two or more signs of tissue hypoperfusion
Mathur et al26	50	53.7	32 (64)	S	25.1/NR/NR	SBP =75.6	SBP <90 mmHg and two or more of the systemic inflammatory response syndrome
Lauzier et al25	23	54.7	14 (60.9)	M	23.2/NR/8.9	MAP <60	Septic shock with MAP <60 mmHg after >1,000 mL crystalloid resuscitation, vasopressors used <12 hours, PAOP ≥12 mmHg, cardiac index ≥3 L/min/m2
Seguin et al24	22	66	17 (77.3)	S	NR/54/10	SBP <90	SBP <90 mmHg; infection; three or more of the systemic inflammatory response syndrome; two or more following criteria: plasma lactate >2 mmol/L or pH <7.3, hypoxemia, urine output <30 mL/hour, platelet count <100,000/mm3, or a decrease of 50% from a previous value or unexplained coagulopathy
Albanese et al23	20	65.5	13 (65)	S	28.5/NR/NR	MAP <60	MAP <60 mmHg and two or more organ dysfunctions
Seguin et al22	22	67.5	12 (54.5)	S	NR/59.5/10	SBP <90	SBP <90 mmHg; infection; three or more of the systemic inflammatory response syndrome; two or more following criteria: plasma lactate >2 mmol/L or pH E02C7.3, hypoxemia, urine output <30 mL/hour, platelet count <100,000/mm3 or a decrease of 50% from a previous value or unexplained coagulopathy
Levy et al21	30	55	21 (70)	S	23.5/NR/NR	MAP <60	After optimal fluid resuscitation and dopamine up to a dose of 20 μg/kg/min, the patients still have the following criteria: MAP <60 mmHg, urine output <30 mL/hour, increased lactate level (>2.5 mmol/L, cardiac index ≥3.5 L/min/m2
Marik and Mohedin18	20	46	11 (55)	S	17.5/NR/NR	MAP <60	After optimal fluid resuscitation, the patients with sepsis still had cardiac index >3.2 L/min/m2 or SVRI <1,200 dyne·s·cm−5·m2 or MAP <60 mmHg
Martin et al20	32	52.5	24 (75)	S	30.5/NR/NR	SBP <90	SBP <90 mmHg, cardiac index >4 L/min/m2, decreased organ perfusion, lactate levels of arterial blood >2.5 mmol/L, and infection
Ruokonen et al19	10	45.1	NR	S	13.3/NR/NR	SBP <90	SBP <90 mmHg with PAOP of 8 mmHg to 12 mmHg, infection

Abbreviations: APACHE II, Acute Physiology and Chronic Health Evaluation II; CVP, central venous pressure; DBP, diastolic blood pressure; M, multicenter trial; MAP, mean arterial pressure; NR, not reported; PAOP, pulmonary artery occlusion pressure; RCT, randomized controlled trial; S, single-center trial; SAPS, simplified acute physiology score; SBP, systolic blood pressure; SOFA, sequential organ failure assessment; SVRI, systemic vascular resistance index; NE, norepinephrine.

All studies evaluated the vasopressor effects in patients with septic shock by using a primary outcome such as survival, hemodynamics, or APACHE II score (Table 2). Vasopressor agents include NE,9,10,18–20,23,25,26,28–35 EN,21,22,24,27,28 VP,10,25,30,31,33 DA,18–20,26,34,35 TP,23,31 PE,9,32 NE+DB,11,21,22 NE+EN,11 TP+NE,9 TP+DB,9 and NE + dopexamine (DX) (Table 2, Figure 2).24 The mortality data from the RCT by De Backer et al7,35 were extracted from their meta-analysis.

Table 2

The interventions of different vasopressors in the included randomized trials

Source	Interventions	Outcome
		Hemodynamic variables	Organ function/metabolic parameters	Mortality	Cardiac adverse events
Mahmoud and Ammar11	NE+DB vs NE+EN	NE+EN is more effective than NE+DB in reversing the abnormalities of cardiovascular parameters	NE+EN group had lower arterial pH and higher serum lactate	No difference	Lower in NE+EN group
Gordon et al30	NE vs VP	VP decreases HR	No difference	No difference	No difference
De Backer et al35	NE vs DA	NR	NR	No difference	NR
Patel et al34	NE vs DA	NR	NR	No difference	Lower in NE group
Gordon et al33	NE vs VP	NR	VP may reduce progression to renal failure	VP decreased 28-day mortality	NR
Jain and Singh32	NE vs PE	PE decreased HR and improved SVI	No differences	No difference	NR
Morelli et al31	NE vs VP vs TP	TP decreased HR	Creatinine increased only in NE group	No difference	No differences
Morelli et al9	NE vs PE	No differences	No differences	No differences	No differences
Morelli et al29	NE vs TP+NE vs TP+DB	TP (or +DB) increased MAP	TP (or +DB) increased urinary output	No differences	NR
Myburgh28	NE vs EN	No difference	No differences	No difference	No differences
Russell et al10	NE vs VP	VP decreases HR	No difference	No difference	No difference
Annane et al27	NE+DB vs EN	No difference	No difference	No difference	No difference
Mathur et al26	NE vs DA	NE improved SBP, HR, CI, and SVRI	NE improved DO2 and urine output	No difference	NR
Lauzier et al25	NE vs VP	NE was required less than VP at 48 hours when MAP was increased equally	VP improved renal function and SOFA score	No difference	No difference
Seguin et al24	NE+DX vs EN	EN increased HR and CO	NE+DX enhanced gastric mucosal blood flow	No difference	NR
Albanese et al23	NE vs TP	TP decreased CI and HR	No difference	No difference	NR
Seguin et al22	NE+DB vs EN	EN increased CI and oxygen transport	EN enhanced gastric mucosal blood flow	No difference	NR
Levy et al21	NE+DB vs EN	No difference	EN enhanced the lactate/pyruvate ratio	No difference	No difference
Marik and Mohedin18	NE vs DA	DA increased MAP by increasing the CI, whereas NE did so by increasing SVRI	NE may improve splanchnic tissue oxygen utilization	No difference	NR
Martin et al20	NE vs DA	NE is more effective than DA in reversing the abnormalities of hyperdynamic septic shock	NE can increase mean perfusion pressure without apparent adverse on renal blood flow	No difference	NR
Ruokonen et al19	NE vs DA	No difference	DA increased splanchnic flow and DO2 but decreased VO2	No difference	NR

Abbreviations: CI, cardiac index; CO, cardiac output; DA, dopamine; DB, dobutamine; DO2, oxygen delivery; DX, dopexamine; EN, epinephrine; HR, heart rate; MAP, mean arterial pressure; NE, norepinephrine; NR, not reported; PE, phenylephrine; SOFA, sequential organ failure assessment; SVI, stroke volume index; SVRI, systemic vascular resistance index; VO2, oxygen consumption; TP, terlipressin; VP, vasopressin; vs, versus; SBP, systolic blood pressure.

Figure 2

Network of eligible comparisons for the multiple-treatment meta-analysis for mortality.

Notes: The width of the lines is proportional to the number of trials comparing each pair of treatments, and the size of each node is proportional to the number of randomized participants (sample size). The network of eligible comparisons for acceptability (dropout rate) analysis is similar.

Abbreviations: DA, dopamine; DB, dobutamine; DX, dopexamine; EN, epinephrine; NE, norepinephrine; PE, phenylephrine; TP, terlipressin; VP, vasopressin.

Risk of bias within studies

Only RCTs were included in the analysis. Sequence of randomized allocation was reported in all but two studies.22,34 Blinding was conducted in nine studies.9–11,20,27,30,32,33,35 The mean Jadad score was 3.3.

Effect of different vasopressor agents on mortality

Mortality in these 21 trials was 50.1% (1,915/3,819). When compared to NE, DA was associated with increased mortality (OR: 1.24, 95% CI: 1.01, 1.53). However, there was no significant difference in mortality in direct or indirect comparisons between other different vasopressor agents and vasopressor combinations (P>0.05) (Figure 3). For the probability of mortality, the possible rank from low to high was NE+DB (area under the curve [AUC]: 0.2648), EN (AUC: 0.3473), TP (AUC: 0.379), NE+EN (AUC: 0.3943), TP+NE (AUC: 0.3967), VP (AUC: 0.4212), TP+DB (AUC: 0.5423), NE (AUC: 0.5752), PE (AUC: 0.6796), NE+DX (AUC: 0.7279), and DA (AUC: 0.7718) (Figures 4 and 5). The tests of inconsistency for the two triangular closed loops were not significant (Figure 6). This meant that direct and indirect estimates had similar effects in the closed loop.15,17

Figure 3

Mortality of different vasopressors in direct comparison and network meta-analysis in terms of mortality.

Notes: Results are the ORs and CIs in the row-defining treatment compared with the ORs and CIs in the column-defining treatment. For mortality, ORs >1 favor the row-defining treatment. Network meta-analysis results are at the bottom-left of the figure, while direct comparison results are at the upper-right of the figure.

Abbreviations: CI, confidence interval; DA, dopamine; DB, dobutamine; DX, dopexamine; EN, epinephrine; NE, norepinephrine; NMA, network meta-analysis; OR, odds ratio; PE, phenylephrine; TP, terlipressin; VP, vasopressin.

Figure 4

Ranking for mortality.

Notes: Ranking indicates the probability to be the most mortality risk treatment, the second best, the third best, and so on, among the vasopressor agents.

Abbreviations: DA, dopamine; DB, dobutamine; DX, dopexamine; EN, epinephrine; NE, norepinephrine; PE, phenylephrine; TP, terlipressin; VP, vasopressin.

Figure 5

The cumulative probability plot.

Notes: Area under the curve indicates the probability to be the most mortality risk treatment, the second best, the third best, and so on, among the vasopressor agents.

Abbreviations: DA, dopamine; DB, dobutamine; DX, dopexamine; EN, epinephrine; NE, norepinephrine; PE, phenylephrine; TP, terlipressin; VP, vasopressin.

Figure 6

Inconsistency for triangular loops.

Notes: acd: norepinephrine, vasopressin, and terlipressin comparison closed loop; afg: norepinephrine, terlipressin + dobutamine, and terlipressin + norepinephrine comparison closed loop. The values are shown as mean (confidence interval of inconsistency estimate). The symbol ■ indicates sample size.

Effect of different vasopressor agents on cardiac adverse events

Included studies compared NE vs DA, NE vs VP, NE vs TP, NE vs PE, TP+NE vs TP+DB, and TP+DB vs EN directly. We performed direct meta-analysis of cardiac adverse events, which mainly consisted of arrhythmias and tachycardia. NE decreased cardiac adverse events significantly compared to DA (Table 3). No significant difference in cardiac adverse events was found between other vasopressor agents and vasopressor combinations.

Table 3

Direct comparison of different vasopressors on cardiac adverse events

	Number of studies	Number of patients	OR (95% CI)	Heterogeneity I2 (P-value)	Test for effect (P-valuea)
NE vs DA	134	252	0.15 (0.05, 0.43)	–	0.0005
NE vs VP	310,25,31	831	1.30 (0.73, 2.32)	0% (0.48)	0.38
NE vs TP	131	30	12.13 (0.59, 248.49)	–	0.11
NE vs PE	19	32	0.47 (0.04, 5.73)	–	0.55
TP+NE vs TP+DB	127	330	0.88 (0.53, 1.45)	–	0.61
TP+DB vs EN	111	60	0.66 (0.18, 2.36)	–	0.52

Note:

Fixed-effect model.

Abbreviations: CI, confidence interval; DA, dopamine; DB, dobutamine; EN, epinephrine; NE, norepinephrine; PE, phenylephrine; TP, terlipressin; VP, vasopressin; vs, versus.

Effect of different vasopressors on hemodynamic and metabolic parameters

Thirteen studies reported that there were significant differences in the effect on hemodynamics,10,11,18,20,22–26,29–32 and eleven studies reported that there were significant differences on metabolic parameters or organ function between vasopressor agents and vasopressor combinations (Table 2).11,18–22,24–26,29,31,33

Four trials with complete data compared the treatment of NE and DA.18–20,26 The results revealed that NE decreased HR (SMD: −2.10; 95% CI: −3.95, −0.25; P=0.03) and cardiac index (SMD: −0.73; 95% CI: −1.14, −0.03; P=0.004) and increased SVRI (SMD: 1.03; 95% CI: 0.61, 1.45; P<0.0001), but there was no significant difference on MAP, oxygen delivery (DO2), oxygen consumption (VO2), and lactate. In contrast, as compared to NE, VP significantly decreased HR (SMD: 0.21; 95% CI: 0.07, 0.34; P=0.003).

Compared to the NE+DB combination, EN did not show a significant difference in HR, MAP, cardiac index, pulmonary MAP, DO2, VO2, and lactate (Table 4). However, the NE+EN combination was more effective than the NE+DB combination in reversing the abnormalities of cardiovascular parameters, and the NE+EN group had significantly higher MAP, HR, CVP, cardiac index, SVRI, ejection fraction, left ventricular end diastolic volume, DO2, lactate, and urine output.11

Table 4

Direct comparison of different vasopressors on hemodynamic and metabolic parameters

	Number of studies	Number of patients	SMD IV (95% CI)	Heterogeneity I2 (P-value)	Test for effect (P-value)
NE vs DA
HR	418–20,26	105	−2.10 (−3.95, −0.25)	91% (<0.0001)	0.03a
MAP	318–20	55	0.64 (−1.09, 2.38)	87% (0.0004)	0.47a
Cardiac index	418–20,26	105	−0.73 (−1.14, −0.03)	43% (0.15)	0.004b
SVRI	418–20,26	105	1.03 (0.61, 1.45)	26% (0.25)	<0.0001b
DO2	418–20,26	105	−0.54 (−1.50, 0.42)	79% (0.003)	0.27a
VO2	418–20,26	105	−0.49 (−1.37, 0.39)	75% (0.008)	0.27a
Lactate	318–20	55	0.01 (−0.53, 0.56)	23% (0.27)	0.96b
NE vs VP
HR	310,25,31	831	0.21 (0.07, 0.34)	0% (0.96)	0.003b
MAP	310,25,31	831	−0.07 (−0.21, 0.07)	0% (0.70)	0.76b
Cardiac index	325,30,31	294	−0.04 (−0.26, 0.19)	0% (0.93)	0.76b
SVRI	225,31	53	0.15 (−0.39, 0.70)	0% (0.91)	0.58b
DO2	225,31	53	−0.06 (−0.62, 0.49)	0% (0.42)	0.82b
VO2	225,31	53	0.03 (−0.52, 0.59)	0% (0.44)	0.91b
Lactate	225,31	53	0.25 (−0.31, 0.80)	0% (0.95)	0.38b
NE+DB vs EN
HR	221,22	52	0.33 (−0.22, 0.89)	49% (0.16)	0.24b
MAP	221,22	52	−0.24 (−0.78, 0.31)	0% (0.99)	0.90b
Cardiac index	221,22	52	−0.04 (−0.59, 0.51)	48% (0.17)	0.90b
MPAP	221,22	52	−0.09 (−0.63, 0.45)	0% (0.71)	0.75b
DO2	221,22	52	−0.19 (−0.74, 0.36)	47% (0.17)	0.50b
VO2	221,22	52	−0.13 (−0.67, 0.42)	0% (0.41)	0.65b
Lactate	221,22	52	−0.11 (−0.66, 0.43)	0% (0.59)	0.69b

Notes:

Random-effects model;

fixed-effect model.

Abbreviations: CI, confidence interval; DA, dopamine; DB, dobutamine; DO2, oxygen delivery; EN, epinephrine; HR, heart rate; IV, inverse variance method; MAP, mean arterial pressure; MPAP, mean pulmonary arterial pressure; NE, norepinephrine; SMD, standardized mean difference; SVRI, systemic vascular resistance index; VO2, oxygen consumption; VP, vasopressin; vs, versus.

Discussion

Twenty-one trials that included 3,819 patients and that compared different vasopressor agents or vasopressor combinations in septic shock were identified and included in this systematic review and NMA of RCTs. The trials’ mean Jadad score was 3.3, which means that they were of high quality. The main results showed that except for the superiority of NE over DA in direct comparison, the mortality of patients treated with any other vasopressor agent or vasopressor combination was not significantly different. NE was also associated with decreased cardiac adverse events, HR, and cardiac index, as well as increased SVRI, as compared to DA.

Our meta-analysis revealed a possible rank of probability of mortality among the eleven vasopressor agents or vasopressor combination; from low to high, they are NE+DB, EN, TP, NE+EN, TP+NE, VP, TP+DB, NE, PE, NE+DX, and DA. However, variations in each RCT’s inclusion criteria may have influenced the probability of mortality. Thus, this ranking should be interpreted with caution.

Our NMA evaluated the vasopressor agents or vasopressor combinations from both direct and indirect comparisons. This approach differs from traditional head-to-head meta-analysis. Some traditional meta-analyses of RCTs have compared only two or three vasopressor agents, such as NE, DA, and VP.10,35 However, other types of comparisons have never been performed. This NMA compared any vasopressor agent or vasopressor combination to others and revealed a possible rank of probability of mortality.15

Three factors support the internal validity of our analysis. First, a rigorous and extensive literature search was conducted, and the number of selected studies was more than any in previous meta-analyses focusing on vasopressor agents and vasopressor combinations for the treatment of septic shock. Second, the selected trials are considered high-quality studies, with a mean Jadad score of 3.3 points. Third, tests of inconsistency for triangular loops were not significant; in other words, the direct and indirect estimates had similar effects. This finding supports that our NMA has adequate homogeneity, which translates to more confidence in support of the results.

Vasopressor therapy is recommended by every major clinical practice guideline when fluid resuscitation fails to maintain adequate blood pressure and organ perfusion. However, different vasopressor agents and vasopressor combinations increase blood pressure through different mechanisms, leading to heterogeneity of physiological effects.37 NE is the first-line vasopressor agent used to treat septic shock (grade 1B)5 and is associated with lower mortality compared to DA.6,7 Although the typical order for the addition of vasopressor agents is NE, epinephrine, VP, DA, and PE,38 the supporting evidence for this order is limited except for the superiority of NE over DA in terms of mortality.6,7 NE supplemented with EN is the second choice in treating septic shock (grade 2B).5 In this meta-analysis, only one study reported NE+EN vs NE+DB.11 The rank of probability of mortality revealed that NE+EN had lower risk than NE. VP is neither recommended nor suggested (grade UG) but can be added to NE with the intent of either raising MAP or decreasing NE dosage.5,38 PE, which is used to stimulate purely α-1 receptors, is recommended when cardiac output is known to be high and the target blood pressure is not achieved (grade 1C).5 No significant difference between PE and other vasopressor agents or vasopressor combinations was found. Similar results were also found in the comparison between other vasopressor agents or vasopressor combinations. Recently, a trial compared the vasopressor effects of NE+DB and NE+EN on the cardiovascular support of patients with septic shock.11 To better evaluate any mortality benefit from the initial vasopressor used, we also compared vasopressor combinations of NE+DB, TP+NE, TP+DB, NE+EN, and NE+DX. The results showed that the vasopressor combination NE+DB had the lowest probability of mortality, and this finding may be supported by the rapid normalization of both gastric–arterial difference (PCO2 gap) and gastric intramu-cosal pH.22 No other vasopressor combination is superior to another in both direct and indirect comparisons.

For cardiac adverse events and hemodynamic and metabolic parameters, we conducted only direct comparisons because the small number of studies failed to form an effective network analysis loop. Our direct meta-analysis revealed that cardiac adverse events, HR, and cardiac index were decreased and SVRI was increased on treatment with NE compared to the results of treatment with DA. These results support the notion that NE may have stronger α-receptor effects, resulting in a greater increase in SVRI and blood pressure as compared to DA.4,39 Even though some studies favored NE as the more effective vasopressor agent to maintain adequate MAP during septic shock, no significant difference in terms of effect on MAP between these two vasopressor agents has ever been detected.20,40 Overall, NE is probably more effective than DA in hemodynamic support for septic shock patients.

A previous trial reported that VP might increase SVRI and decrease cardiac index compared to baseline, while NE did not.25 Meta-analysis that included two trials failed to find any significant difference in cardiac adverse events as well as hemodynamic and metabolic parameters between NE and VP.

Statistically, with 80% power and two-sided alpha level of 0.04, to detect a 15% relative difference in 28-day mortality rate, at least 765 subjects in each group were needed.35 In the present meta-analysis, only “NE vs VP” (n=1,799) and “NE vs DA” (n=1,408) comparisons had potentially adequate sample size.

Limitations

Our analysis has many limitations. First, only English language articles were included in this study, which may have affected the findings due to selection bias. Second, although 21 trials were included in this study, the actual sample size population in specific comparisons was small, and the risk of false attribution of positive effect from pooling small trials is well known. Moreover, differences in each RCT’s inclusion criteria may have influenced the probability of mortality. Additionally, publication bias analysis could not be conducted. Hence, we do not think that these results constitute a reason to change clinical practice, but rather, they support the need for further investigations.

Conclusion

In terms of survival, NE may be superior to DA. Otherwise, there is insufficient evidence to suggest that any other vasopressor agent or vasopressor combination is superior to another. When compared to DA, NE is associated with decreased cardiac adverse events, HR, and cardiac index, as well as increased SVRI. The effects of vasopressor agents or vasopressor combinations on patients with septic shock require further investigation by larger-scale RCTs.