Secondary Logo

Journal Logo

Clinical Science Aspects

Norepinephrine vs Vasopressin: Which Vasopressor Should Be Discontinued First in Septic Shock? A Meta-Analysis

Wu, Zongsheng∗,†; Zhang, Shi; Xu, Jingyuan; Xie, Jianfeng; Huang, Lili; Huang, Yingzi; Yang, Yi; Qiu, Haibo

Author Information
doi: 10.1097/SHK.0000000000001345



Septic shock, which is the most severe subset of sepsis, accounts for 10% of all ICU admissions and has an associated mortality rate of 40% to 60%. In the case of life-threatening hypotension of septic shock, vasopressor therapy should be initiated immediately (1, 2). Vasoactive agents, such as norepinephrine (NE) and vasopressin (VP), are used widely in this situation. The Surviving Sepsis Campaign (SSC) guidelines in 2016 recommend NE as the initial vasoactive agent of choice based upon moderate evidence (2). However, NE monotherapy often fails to reverse shock. In the landmark, vasopressin in Septic Shock Trial (VASST), vasopressin deficiency was observed in the early phase of septic shock (3). Moreover, to achieve the target mean arterial pressure (MAP) in 1 h (recommendation of the SSC Bundle: 2018 Update), concomitant NE and VP therapy was shown to work faster than NE monotherapy in this period (4). Therefore, concomitant NE and VP infusions are commonly used in the clinic.

With improvements in treatment and hemodynamics, clinicians should decide which vasoactive agent to discontinue first once a patient with septic shock begins to improve. To date, there is lack of guidance regarding the appropriate discontinuation order of these agents. In an early retrospective study performed by Bauer et al. (5), the cessation of VP before NE resulted in a greater incidence of clinically significant hypotension. Several other studies argued that tapering NE rather than VP may be associated with a higher incidence of hypotension (6, 7). Nevertheless, a recent large cohort study concluded that there was no significant difference in the incidence of hemodynamic instability based on the discontinuation sequence of NE and VP (8). Therefore, the optimal approach to weaning vasopressors in septic shock remains controversial. The present meta-analysis was performed to investigate the effect of vasopressor discontinuation order on the occurrence of hypotension and mortality in adult patients with septic shock.


This study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses: the PRISMA statement (9).

Search strategy and selection criteria

A literature review of all pertinent English language studies was undertaken in PubMed, Cochrane Central Register, and EMBASE by two authors, independently, from inception through October 18, 2018. The keywords included vasopressin, norepinephrine, vasopressor, vasoactive agents, hypotension, sepsis, shock, septic shock, outcome, mortality, and adverse. In addition, reference lists of included studies were manually searched for relevant studies.

Title and abstract screening was conducted for all relevant records based on the electronic literature search. A full-text review was performed for potentially relevant records. Randomized or observational reports were eligible to be included in the meta-analysis if they enrolled adult patients (≥18 yr old), included the septic shock population, and included patients on concomitant vasopressin and norepinephrine and reported the discontinuation sequence of two vasoactive agents. Abstracts from conference proceedings were also included. Pediatric studies, case reports/series, and editorial/comments were excluded.

Study quality

The risk of bias was assessed for each outcome in all included studies using the Cochrane Systematic Review Handbook for Randomized Controlled Studies (10) and the Newcastle-Ottawa scale for Nonrandomized Controlled Studies (11). Each study was scored from 0 to 9 based upon the eight criteria covering the selection of the cohort, comparability of groups, and the outcome. Discrepancies between the two authors were resolved by consensus.

Data extraction and synthesis

The data collected from each study included general information (author, year, and study design), characteristics of the participants (including gender, age, and inclusion and exclusion criteria), outcome measurements, with primary outcome determined as hypotension and secondary outcomes as clinical outcomes including overall mortality (when >1 value for mortality was provided by the article, the mortality for the longest complete follow-up was preferentially used in the meta-analysis.), ICU mortality, ICU length of stay (LOS), time to hypotension, and intervention after hypotension (restarting of the vasopressor or an increase in the remaining agent).

Dichotomous variables were expressed as counts and proportions. Means and SD were used to describe normally distributed continuous variables. Because the ICU LOS and time to hypotension data were often not normally distributed, certain studies reported the ICU LOS and time to hypotension data using the median and the first and third quartiles. To include these data in the present study, we estimated the sample mean and SD based on the method presented by Wan et al. (12) and Luo et al. (13). Notably, this method is based on assumption that the data are normally distributed, which we know is not the case.

Sensitivity and subgroup analysis

Sensitivity analyses were performed by removing each study individually to determine whether an individual report had a higher contribution to the heterogeneity or overall effect estimate, analyzing published studies separately from abstracts, and including only high-quality reports (NOS ≥ 7).

Subgroup analyses were performed to examine the following parameters: high corticosteroid usage rate (≥75%) vs low corticosteroid usage rate (<75%); and studies enrolling patients with an average age ≥65 years vs <65 years.

Statistical analysis

The data retrieved from the relevant articles were computerized and analyzed by Review Manager V.5.3 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark). The standard deviation of mean differences with 95% CI for continuous outcomes (ICU LOS), and the OR with 95% CI for dichotomous outcomes (hypotension, mortality, and intervention after hypotension) were used to estimate the pooled effects. The random-effects model was used for better accommodation of heterogeneity. A sensitivity analysis was performed by removing each study individually using Stata version 14.0 (StataCorp, College Station, TX).

Heterogeneity was tested using the Cochran Q statistic (P < 0.1) and quantified with the I2 statistic, with a range of 0% to 30% representing no or mild heterogeneity, 30% to 60% representing moderate heterogeneity and >60% representing high heterogeneity.

A trial sequential analysis (TSA) was performed to estimate the optimal sample size to reach a plausible conclusion on the optimal vasopressor discontinuation sequence in the case of septic shock. Trial Sequential Analysis V. beta (Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet, Copenhagen, Denmark, available from was used. Statistical significance was set at a two-tailed P value of 0.05.


Study search and selection

The flow chart for the selection process and detailed identification is presented in Figure 1. A total of 601 publications were identified through the initial search of the databases. After screening the titles and abstracts, 16 studies were selected for a full-text review. Ultimately, six published studies and two conference abstracts were included in the meta-analysis.

Fig. 1
Fig. 1:
Literature search flow diagram.

Study characteristics and quality

Table 1 shows the characteristics of the trials included in the meta-analysis. The number of participants across trials ranged from 50 to 585. The years of publication ranged from 2010 to 2018. All six full-text publications were single-center trials, and 52.36% of the subjects were male. Among these publications, five reports were retrospective cohort studies and were conducted in the United States; only one report was a prospective randomized controlled trial (RCT), which was conducted in Korea.

Table 1
Table 1:
Characteristics of included studies

In total, hypotension developed when NE was discontinued first for 38.87% (288/741) of the patients, whereas hypotension was reported in 53.9% of the patients (228/423) when VP was discontinued first. In the quality assessment, the range of Newcastle-Ottawa score (NOS) scores was between 6 and 9 (maximum of 9), and only one RCT was judged to have a low risk of bias (Supplemental Table 1a and b,

Syntheses of results

A random effects model was used for all analyses, as described in the methods. Compared with discontinuing VP first, the incidence of hypotension was significantly lower when NE was discontinued first (OR 0.30, 95% CI 0.10 to 0.86, P = 0.02) (Fig. 2). The time from tapering the first vasopressor to the occurrence of hypotension, in hours, was analyzed. Among four studies reporting this outcome, the time to hypotension was shorter when VP was discontinued first, but the difference was not significant (mean difference −1.33, 95% CI −2.71 to 0.05, P = 0.06) (Supplementary Fig. 1,

Fig. 2
Fig. 2:
The effect of different vasopressor discontinuation order on occurrence of hypotension in patients with septic shock.

We compared the effect of the discontinuation order on mortality in septic shock and detected no significant difference in either overall mortality or ICU mortality between the two groups (OR 1.28, 95% CI 0.77 to 2.10, P = 0.34; OR 0.99, 95% CI 0.74 to 1.34, P = 0.96) (Fig. 3).

Fig. 3
Fig. 3:
The effect of different vasopressor discontinuation order on overall mortality (A) and ICU mortality (B) in patients with septic shock.

Furthermore, the ICU LOS was also evaluated in five studies, and no statistical significance was detected between the two orders of weaning vasopressors (mean difference, 1.35, 95% CI −2.05 to 4.74, P = 0.44). However, the ICU LOS was reported as the median and interquartile ranges in three studies. The removal of the data from these studies from the overall analysis reduced the heterogeneity to 0% but did not alter the significance of the result (mean difference, 1.82, 95% CI −0.67 to 4.32, P = 0.15) (Supplementary Fig. 2,

Intervention after the occurrence of hypotension was reported in five studies. Compared to the frequency of restarting VP in patients for whom VP was discontinued first, patients for whom NE was discontinued first had NE restarted more frequently (OR 6.10, 95% CI 1.41 to 26.44, P = 0.02). On the other hand, compared to patients for whom NE was discontinued first, patients who had VP discontinued first more often required an increase in the dose of the remaining vasoactive agent (NE) (OR 0.05, 95% CI 0.01 to 0.34, P = 0.002) (Supplementary Fig. 3,

Sensitivity analysis

In the sensitivity analysis evaluating the removal of individual studies, only the removal of the prospective study by Jeon et al. (7) did not result in a significant change in the overall OR. Moreover, in the sensitivity analyses testing published studies vs abstracts, the OR for published articles was elevated (Supplementary Tab. 2 and Fig. 5, Sensitivity analyses using high-quality reports with an NOS greater than or equal to 7 and one prospective, double-blind, RCT report demonstrated that no significant difference was noted in the incidence of hypotension based on the discontinuation order of NE and VP (OR 0.42, 95% CI 0.10 to 1.79, P = 0.24) (Supplementary Fig. 4, However, the high level of heterogeneity remained substantial after multiple sensitivity analyses were conducted.

Subgroup analysis

Five studies reported the average age of the trial participants, and six studies reported the rate of corticosteroid application. We attempted to divide the studies according to the patients’ average age (<65 years vs ≥65 years) and corticosteroid usage rate (<75% vs ≥75%) and, ultimately, we detected no significant heterogeneity between the subgroups based on a test of interaction (P > 0.05). However, it should be noted that VP discontinuation prior to the cessation of NE was associated with hypotension in the group with a low corticosteroid usage rate (OR 0.18, 95% CI 0.04 to 0.78, P = 0.02) (Fig. 4).

Fig. 4
Fig. 4:
The effect of VP vasopressin discontinued first in the different corticosteroids using rate for patients with septic shock.

Trial sequential analysis

We used a random effects model for all overall low risk of bias trials included in the primary analyses. The TSA was performed with a diversity-adjusted information size that was calculated using a two-sided α of 0.05, a power of 80%, and a control event rate of 50%. This calculation indicated a required information size of 11 821 patients for identifying the optimal order for discontinuing vasoactive agents in septic shock. The Lan-DeMets sequential monitoring boundary constructed by the optimal information size did not cross, indicating that the cumulative evidence was not conclusive and reliable (Fig. 5).

Fig. 5
Fig. 5:
Trial sequential analysis. Before the required information size (RIS = 11 821) is achieved, the assumed effect of NE discontinued first is in fact absent when the Z-curve did not surpasses the Lan-DeMets’ trial sequential monitoring boundary for benefit.


Septic shock remains one of the major causes of morbidity and mortality in the critically ill. Accumulated evidence demonstrates that targeting and maintaining the goal MAP along with the early initiation of vasoactive agents in patients with septic shock is associated with reduced mortality rates (14, 15). Recently, the 2018 update of the SSC Bundle highlighted that hemodynamic stability and adequate perfusion to vital organs should be achieved within 1 h (4). It is well known that relative VP deficiency is a physiologic consequence that occurs early hours of septic shock (16). Meanwhile, in the setting of refractory septic shock, VP can be utilized to enhance the MAP in patients who are receiving NE or to decrease catecholamine requirements (17, 18). Furthermore, in one prospective, open-label study, concomitantly treating patients with VP and NE achieved and maintained an MAP of 65 mmHg faster than treatment with initial NE alone (19). Based on these findings, there is increasing interest in adding VP early as an adjunctive agent to NE. However, the SSC guidelines do not currently advise clinicians as to which vasoactive agent to discontinue first once the patient's septic shock begins to resolve. Although several authors have published studies on this topic already (5–8, 20–23), the optimal approach to discontinuing vasopressors in patients with septic shock remains controversial.

To our knowledge, this is the first meta-analysis demonstrating that discontinuing VP before NE leads to an increased incidence of hypotension but is not associated with worse patient outcomes, such as mortality (overall and ICU) or ICU LOS. There may be several possible explanations for our findings. First, plasma VP deficiency is commonly observed among vasodilatory septic shock patients, and this state may be prolonged (24). The VASST trial highlighted that circulating levels of VP decrease within 24 h following the discontinuation of exogenous VP (3). If patients benefit from VP, it is logical that they would experience negative hemodynamic effects when VP was discontinued first (25). Second, VP has a vasoconstrictive effect, although the VP1a receptor is independent of α-stimulation and cannot be substituted by catecholamine (26). Additionally, the dosage or equivalent dosage of VP was shown to be higher when VP was discontinued first compared with when NE was discontinued first, according to two studies (22, 23). Thus, it will be better to adopt a fixed dose of VP in future prospective trails to determine the effect of VP weaning. Finally, although the available evidence does not support the hypothesis that septic shock patients have significant survival improvement if VP is the final vasopressor discontinued, the CI indicated a potential clinically important benefit for VP. Similar results was also obtained in the VASST trial, which compared NE with NE plus VP and detected no significant difference in 28-day mortality (3). Therefore, larger trials may be warranted to further evaluate this effect.

We analyzed the time to hypotension onset after the discontinuation of the vasoactive agent and found that patients for whom VP was discontinued first appeared to have a shorter period to hypotension onset, but the difference was not significant. This finding may be due to rapid VP clearance, since the half-life of arginine vasopressin in patients with vasodilatory shock is less than 10 min (8).

In addition, it is interesting to note that physicians prefer increasing the NE dose to restarting VP in the case when VP is discontinued first. In contrast, in patients for whom NE is discontinued first, NE was restarted more frequently compared with the resumption of VP in patients for whom VP was discontinued first. SSC guidelines recommend NE as the first-line vasopressor and VP as a second-line supplementation for shock. Another reason for this preference may be that the price of VP has increased exponentially in the United States since 2014 (27). Moreover, Wu et al. (28) reported that a high dosage of NE (50 μg/min) alone appeared to be safe and did not affect the time required to reach the target MAP.

Recent studies have suggested that there may be a potential interaction between VP and corticosteroids (29). For instance, VP can bind to V1b receptors in the anterior pituitary, leading to the release of adrenocorticotropin hormone, and corticosteroids have been shown to restore the cytokine-mediated downregulation of VP receptors (30). Furthermore, the combination of vasopressin and steroids was shown to result in a lower mortality compared with the combination of norepinephrine plus steroids in a less severe septic shock (3, 31). In a subgroup analysis, we divided the studies into two subgroups according to a predefined corticosteroids usage rate of 75%. A higher incidence of hypotension was observed when VP was discontinued first in studies in which patients received corticosteroid therapy percentage below 75%. It is worth noting that only the study by Jeon et al. (7) indicated that patients for whom NE was discontinued first were more likely to develop clinically significant hypotension in septic shock, and almost every patient in this study received corticosteroid treatment (96%, 72/78). These results indicate that corticosteroids may alleviate the effect of VP withdrawal and corroborated the concept that corticosteroids can spare VP requirement. In clinical practice, since VP is second choice for septic shock and with high price (2, 27), the majority of physicians may prefer to discontinue VP first. Our study suggested that VP discontinuation first may lead to higher incidence of hypotension. On the contrary, NE discontinued first maybe a better choice for septic shock patients who treated with concomitant VP and NE. It should be noticed that if shock patient received corticosteroids therapy already, the adverse effects of VP discontinued first will be reduced.

We also performed several predefined sensitivity analyses to confirm the robustness of our findings. In a sensitivity analysis evaluating the removal of individual studies, no single study resulted in a significant change in the heterogeneity. Moreover, the removal of the two abstracts or the low-quality studies from the overall analysis did not alter the heterogeneity.

There are several limitations in this review. Every published article included in this meta-analysis was single-center study and was conducted in the United States. The largest sample size was 585 participants. In addition, five of the six studies were retrospective analyses. Therefore, practice variation across the included studies may have contributed to the heterogeneity. Last but not least, we chose to present the results of conference abstracts to examine the impact on the overall OR and compare the results. Conference abstracts have the following limitations: they may have been edited before the final presentation, they are not always peer-reviewed, and some information may be missing.


Discontinuing VP first may lead to a higher incidence of hypotension but has not been associated with poor outcomes in septic shock patient who receives concomitant VP and NE therapy. The use of corticosteroids may mitigate this effect. However, the TSA indicated a lack of firm evidence for these results. A multicenter, prospective, RCT is warranted to confirm these findings.


1. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, Bellomo R, Bernard GR, Chiche JD, Coopersmith CM, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA 315:801–810, 2016.
2. Rhodes A, Evans LE, Alhazzani W, Levy MM, Antonelli M, Ferrer R, Kumar A, Sevransky JE, Sprung CL, Nunnally ME, et al. Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock: 2016. Intensive Care Med 43:304–377, 2017.
3. Russell JA, Walley KR, Singer J, Gordon AC, Hebert PC, Cooper DJ, Holmes CL, Mehta S, Granton JT, Storms MM, et al. Vasopressin versus norepinephrine infusion in patients with septic shock. N Engl J Med 358:877–887, 2008.
4. Levy MM, Evans LE, Rhodes A. The Surviving Sepsis Campaign Bundle: 2018 update. Intensive Care Med 44:925–928, 2018.
5. Bauer SR, Aloi JJ, Ahrens CL, Yeh JY, Culver DA, Reddy AJ. Discontinuation of vasopressin before norepinephrine increases the incidence of hypotension in patients recovering from septic shock: a retrospective cohort study. J Crit Care 25 (2):362.e7–362.e11, 2010.
6. Hammond DA, McCain K, Painter JT, Clem OA, Cullen J, Brotherton AL, Chopra D, Meena N. Discontinuation of vasopressin before norepinephrine in the recovery phase of septic shock. J Intensive Care Med 34 (10):805–810, 2019.
7. Jeon K, Song JU, Chung CR, Yang JH, Suh GY. Incidence of hypotension according to the discontinuation order of vasopressors in the management of septic shock: a prospective randomized trial (DOVSS). Crit Care 22 (1):131, 2018.
8. Sacha GL, Lam SW, Duggal A, Torbic H, Reddy AJ, Bauer SR. Hypotension risk based on vasoactive agent discontinuation order in patients in the recovery phase of septic shock. Pharmacotherapy 38:319–326, 2018.
9. Moher D, Liberati A, Tetzlaff J, Altman DG. Group P. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ 339:b2535, 2009.
10. Higgins JPT: Cochrane handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]. The Cochrane Collaboration 2011 London, UK. Available at: Accessed October 20, 2018.
11. Wells GA, O’Connell D, Peterson J, Welch V, Losos M, Tugwell P. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2013. Available at: Accessed October 20, 2018.
12. Wan X, Wang W, Liu J, Tong T. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol 14:135, 2014.
13. Luo D, Wan X, Liu J, Tong T. Optimally estimating the sample mean from the sample size, median, mid-range, and/or mid-quartile range. Stat Methods Med Res 27:1785–1805, 2018.
14. Dunser MW, Ruokonen E, Pettila V, Ulmer H, Torgersen C, Schmittinger CA, Jakob S, Takala J. Association of arterial blood pressure and vasopressor load with septic shock mortality: a post hoc analysis of a multicenter trial. Crit Care 13:R181, 2009.
15. Beck V, Chateau D, Bryson GL, Pisipati A, Zanotti S, Parrillo Je, Kumar A. Cooperative Antimicrobial Therapy of Septic Shock Database Research(CATSSDR). Timing of vasopressor initiation and mortality in septic shock: a cohort study. Crit Care 18:R97, 2014.
16. Landry DW, Levin HR, Gallant EM, Ashton RC, Seo S, D’Alessandro D, Oz MC, Oliver JA. Vasopressin deficiency contributes to the vasodilation of septic shock. Circulation 95:1122–1125, 1997.
17. Malay MB, Ashton RC, Landry DW, Townsend RN. Low-dose vasopressin in the treatment of vasodilatory septic shock. J Trauma 47:699–703, 1999.
18. Tsuneyoshi I, Yamada H, Kakihana Y, Nakamura M, Nakano Y, Boyle WA 3rd. Hemodynamic and metabolic effects of low-dose vasopressin infusions in vasodilatory septic shock. Crit Care Med 29:487–493, 2001.
19. Hammond DA, Ficek OA, Painter JT, McCain K, Cullen J, Brotherton AL, Kakkera K, Chopra D, Meena N. Prospective open-label trial of early concomitant vasopressin and norepinephrine therapy versus initial norepinephrine monotherapy in septic shock. Pharmacotherapy 38:531–538, 2018.
20. Bredhold B, Allen L, Callison J, Heidel E, Hamilton L. Impact of the sequence of norepinephrine and vasopressin discontinuation post-septic shock. Crit Care Med 46: (Supplement 1): 732, 2018.
21. Curtis J. Discontinuation of vasopressors in septic shock patients weaning from norepinephrine and vasopressin. Crit Care Med 44: (12, Supplement 1): 422, 2016.
22. Bissell BD, Magee C, Moran P, Bastin ML, Flannery AH. Hemodynamic instability secondary to vasopressin withdrawal in septic shock. J Intensive Care Med 34 (9):761–765, 2019.
23. Musallam N, Altshuler D, Merchan C, Zakhary B, Aberle C, Papadopoulos J. Evaluating vasopressor discontinuation strategies in patients with septic shock on concomitant norepinephrine and vasopressin infusions. Ann Pharmacother 52:733–739, 2018.
24. Sharshar T, Blanchard A, Paillard M, Raphael JC, Gajdos P, Annane D. Circulating vasopressin levels in septic shock. Crit Care Med 31:1752–1758, 2003.
25. Leone M, Boyle WA. Decreased vasopressin responsiveness in vasodilatory septic shock-like conditions. Crit Care Med 34:1126–1130, 2006.
26. Micek ST, Shah P, Hollands JM, Shah RA, Shannon WD, Kollef MH. Addition of vasopressin to norepinephrine as independent predictor of mortality in patients with refractory septic shock: an observational study. Surg Infect (Larchmt) 8:189–200, 2007.
27. Curtis N, Corapi J, Roberts R, Devlin JW. Rebranding of generic parenteral vasopressin: effect on clinician practices and perceptions. Am J Health Syst Pharm 74:105–106, 2017.
28. Wu JY, Stollings JL, Wheeler AP, Semler MW, Rice TW. Efficacy and outcomes after vasopressin guideline implementation in septic shock. Ann Pharmacother 51:13–20, 2017.
29. Gordon AC, Mason AJ, Thirunavukkarasu N, Perkins GD, Cecconi M, Cepkova M, Pogson DG, Aya HD, Anjum A, Frazier GJ, et al. Effect of Early Vasopressin vs Norepinephrine on Kidney Failure in Patients With Septic Shock: The VANISH Randomized Clinical Trial. JAMA 316:509–518, 2016.
30. Vincent JL, Post EH. Sepsis: Vasopressin: a first-line agent for septic shock? Nat Rev Nephrol 12:718–719, 2016.
31. Gordon AC, Russell JA, Walley KR, Singer J, Ayers D, Storms MM, Holmes CL, Hebert PC, Cooper DJ, Mehta S, et al. The effects of vasopressin on acute kidney injury in septic shock. Intensive Care Med 36:83–91, 2010.

Discontinuation; hypotension; norepinephrine; septic shock; vasopressin

Supplemental Digital Content

Copyright © 2019 by the Shock Society