Bench-to-bedside review: Vasopressin in the management of septic shock

This review of vasopressin in septic shock differs from previous reviews by providing more information on the physiology and pathophysiology of vasopressin and vasopressin receptors, particularly because of recent interest in more specific AVPR1a agonists and new information from the Vasopressin and Septic Shock Trial (VASST), a randomized trial of vasopressin versus norepinephrine in septic shock. Relevant literature regarding vasopressin and other AVPR1a agonists was reviewed and synthesized. Vasopressin, a key stress hormone in response to hypotension, stimulates a family of receptors: AVPR1a, AVPR1b, AVPR2, oxytocin receptors and purinergic receptors. Rationales for use of vasopressin in septic shock are as follows: first, a deficiency of vasopressin in septic shock; second, low-dose vasopressin infusion improves blood pressure, decreases requirements for norepinephrine and improves renal function; and third, a recent randomized, controlled, concealed trial of vasopressin versus norepinephrine (VASST) suggests low-dose vasopressin may decrease mortality of less severe septic shock. Previous clinical studies of vasopressin in septic shock were small or not controlled. There was no difference in 28-day mortality between vasopressin-treated versus norepinephrine-treated patients (35% versus 39%, respectively) in VASST. There was potential benefit in the prospectively defined stratum of patients with less severe septic shock (5 to 14 μg/minute norepinephrine at randomization): vasopressin may have lowered mortality compared with norepinephrine (26% versus 36%, respectively, P = 0.04 within stratum). The result was robust: vasopressin also decreased mortality (compared with norepinephrine) if less severe septic shock was defined by the lowest quartile of arterial lactate or by use of one (versus more than one) vasopressor at baseline. Other investigators found greater hemodynamic effects of higher dose of vasopressin (0.06 units/minute) but also unique adverse effects (elevated liver enzymes and serum bilirubin). Use of higher dose vasopressin requires further evaluation of efficacy and safety. There are very few studies of interactions of therapies in critical care - or septic shock - and effects on mortality. Therefore, the interaction of vasopressin infusion, corticosteroid treatment and mortality of septic shock was evaluated in VASST. Low-dose vasopressin infusion plus corticosteroids significantly decreased 28-day mortality compared with corticosteroids plus norepinephrine (44% versus 35%, respectively, P = 0.03; P = 0.008 interaction statistic). Prospective randomized controlled trials would be necessary to confirm this interesting interaction. In conclusion, low-dose vasopressin may be effective in patients who have less severe septic shock already receiving norepinephrine (such as patients with modest norepinephrine infusion (5 to 15 μg/minute) or low serum lactate levels). The interaction of vasopressin infusion and corticosteroid treatment in septic shock requires further study.

AVPR1b (or V3 receptor) is expressed in the anterior pituitary gland and hippocampus. Stimulation of AVPR1b by vasopressin releases adrenocorticotropic hormone (ACTH) because vasopressin fl ows from the posterior pituitary through pituitary portal capillaries to bind to the AVPR1b on corticotrophic cells of the anterior pituitary. Vasopressin thus interacts with the corticosteroid axis in response to stresses such as hypotension [13,14]. Vasopressin and corticotrophin-releasing hormone stimulate diff erent signaling systems and have synergistic eff ects on release of ACTH. AVPR1b knockout mice have an impaired stress response because of blunted ACTH response [15]. In contrast, overexpression of AVPR1b has been asso ciated with pituitary adenomas and ectopic ACTH syndrome [16].
AVPR2 (V2 receptor) is expressed in the renal collecting duct. Vasopressin is a trophic factor of the ascending limb of Henle and thereby creates the gradient that mediates vasopressin's antidiuretic eff ect. AVPR2 stimulation increases retention of water (antidiuretic activity) by increasing cyclic AMP, which causes movement of aquaporin-2 water channels from cytoplasm to the apical membrane of collecting duct cells. If there is vasopressin defi ciency, aquaporin-2 channels internalize from the apical membrane to subapical vesicles so that there is no active water reabsorption. Absence of vasopressin also results in decreased synthesis of aquaporin-2, further impairing water reabsorption. Th e V2 receptor also mediates vasodilation by stimulating the nitric oxide pathway [17,18].
Over 150 mutations of the AVPR2 receptor have been reported associated with nephrogenic diabetes insipidus. Stimulation of the AVPR2 receptor increases release of von Willebrand factor, von Willebrand factor multimers, and risk of clotting. Stimulation of AVPR2 is therefore the mechanism for increased coagulation in response to desmopressin. Interestingly, von Willebrand factor levels are increased in sepsis [19,20].
Sepsis may downregulate vasopressin receptors. Expression of AVPR1a is downregulated in studies of models of sepsis, thereby limiting smooth muscle contraction and increase of blood pressure [10,11,21], AVPR2 receptors [22] and AVPR3 receptors [23]. Downregulation AVPR1a in the lung, liver, kidney and heart is probably caused by increased TNFα, IL-1β, IL-6 and IFNγ [10,21]. Th e magnitude of downregulation of AVPR1a may be large because continuous infusion of lipopolysaccharide decreased AVPR1A expression by 43% [11]. Models of sepsis also downregulate AVPR1b and AVPR2. Lipopolysaccharide decreases mRNA levels of the AVPR1b receptor in the pituitary [23] and downregulates AVPR2 expression in rat renal medulla [22].
Vasopressin also simulates oxytocin receptors to cause vasodilation [24,25]. Stimulation of oxytocin receptors increases nitric oxide and thus causes vasodilation [26]. Stimulation of oxytocin receptors also increases release of atrial natriuretic peptide [27,28]. Th e oxytocin receptor is downregulated in the heart in response to ischemiareperfusion injury; this downregulation of the receptor may explain the potentially detrimental eff ects of vasopressin in myocardial ischemia-reperfusion, including increased mortality and decreased myocardial contractility [26].
Purinergic receptors are also stimulated by vasopressin [27]. At low doses, vasopressin acts on purinergic receptors that mediate endothelial vasodilation; this eff ect is reversed when high-dose vasopressin stimulates the AVPR1a receptor to mediate vasoconstriction.
Th e vasopressin neurons of the hypothalamus and neurohypophysis are innervated by a variety of pathways that carry information regarding, fi rstly, blood volume/ blood pressure and, secondly, osmolality. Purinergic recep tors are also important components of the stimulation of release of vasopressin. ATP is a neuro transmitter that signals through the purinergic P2 receptors (P2X2 [28], P2X3 [29,30] and P2Y1 [31,32]) to increase cytosolic free calcium [33], which then stimulates release of vasopressin from the neuro hypo physis [34][35][36][37]. Th e P2X5 receptor is expressed and co-localized with vasopressin on paraventricular and supra optic nuclei of the hypothalamus, suggesting that P2X5 is also important in the stimulation of vasopressin release [38]. Th e stimulatory eff ects of ATP on vasopressin release are synergized (that is, an increased amount and duration of vasopressin release) by co-stimulation with norepinephrine [39] acting as neurotransmitters via P2X [36,39,40] and P2Y1 [41] receptors. Th is synergistic action of ATP and norepinephrine in the neurohypophysis may be impor tant for the sustained release of vasopressin during hypo tension [36]. Th e stimulation of vasopressin release by ATP is terminated by adenosine, a metabolite of ATP, and is thus a natural negative feedback mechanism of vasopressin release [42].
ATP also stimulates P2Y1 and P2Y2 receptors of the convoluted tubule of the nephron to modulate vasopressin antidiuretic eff ects [43,44]. Indeed, P2Y2 knockout mice have salt-resistant hypertension [45]. Furthermore, the cardiac eff ects of vasopressin are mediated at least in part through P2X and P2Y receptors [46].

Synthesis, release and clearance of vasopressin
Vasopressin is synthesized and released into the systemic circulation from the posterior pituitary gland ( Figure 2). Vasopressin, like many hormones, is synthesized as a prohormone and then is cleaved to form the mature active hormone. Serum levels of vasopressin -a nonapeptide -represent the interactions of the synthesis, release, and metabolism of vasopressin. Synthesis of preprovasopressin occurs in neurophypophyseal neurons (also known as magnocellular neurons) of paraventricular and supraoptic nuclei of the hypothalamus. Provasopressin is packaged in neurosecretory granules and transported along the suprahypophyseal tract to the posterior pituitary. Subsequently, there is conversion to provasopressin followed by conversion of provasopressin by subtilisin-like proprotein convertase (SPC3) to vasopressin.
Vasopressin is metabolized by several promiscuous vasopressinases such as LNPEP and IRAP. LNPEP metabolizes vasopressin and also clears oxytocin (a vasodilator), GLUT4 (which modulates cellular uptake of glucose) and angiotensin-converting enzyme 4 (which converts angiotensinogen to the potent vasoconstrictor angiotensin), indicating that variations in levels and function of LNPEP could have complex eff ects on hemodynamics and blood pressure in sepsis.
Sepsis initially increases vasopressin levels by 20-fold to 200-fold to supraphysiologic levels by several mechanisms. Early studies by Wilson and colleagues in endotoxin-induced shock and Escherichia coli-induced shock found that vasopressin levels were >500 pg/ml in dogs and >300 pg/ml in baboons within 15 minutes of initiation of sepsis [47]. Th is increase of vasopressin levels occurred when cardiac output was decreased but before hypotension, suggesting that stimuli other than hypotension -such as increased levels of mediators of infl ammation and decreased cardiac output -stimulated the increase of vasopressin levels. Furthermore, endotoxin stimulates the release of vasopressin into hypophyseal portal blood. Hypotension and hypernatremia stimulate release of stored vaso pressin (immediate) and synthesis of vasopressin (delayed). Vasopressin stores are thus ample and available for instantaneous release from the posterior pituitary in response to sepsis, septic shock and hypotension.
Vasopressin levels increase early in septic shock because hypotension is the most potent stimulus for Provasopressin is synthesized in neurohypophyseal neurons of the paraventricular nuclei (PVN) and supraoptic nuclei (SON) of the hypothalamus. Provasopressin is transported along the supraoptic neurohypophyseal tract in neurosecretory granules to the posterior pituitary gland. Provasopressin is cleaved by enzymes to vasopressin during transport to the posterior pituitary gland. Hypotension and hyperosmolality stimulate release of stored vasopressin (immediate) and synthesis (delayed) of vasopressin. In response to hypotension, vasopressin is released into the systemic circulation and also into the pituitary portal capillaries, fl ows to the anterior pituitary, binds to the AVPR1b receptor of corticotrophs, and stimulates release of adrenocorticotropic hormone. MNPO, median pre-optic nucleus.
release and synthesis of vasopressin. Indeed, if an animal is challenged with confl icting signals to vasopressin release (such as hypotension and hyponatremia), the animal will increase vasopressin. After the initial release of vasopressin, which increases vasopressin levels in septic shock, vasopressin levels then decline rapidlybecause of depletion of stored vasopressin -to levels that are inappropriately low (compared with similarly hypotensive patients who have cardiogenic shock [48,49]). Th e decrease in vasopressin levels is initially due to the depletion of vasopressin in the post-hypophyseal centers, but vasopressin levels remain extremely, inappropriately low [50,51], suggesting that there is also a sustained impairment of synthesis and release of vasopressin. Indeed, consistent with these earlier studies, vasopressin levels remained extremely low for up to 7 days after onset of septic shock in the Vasopressin and Septic Shock Trial (VASST) (Figure 3) [52]. Few endocrine systems are so rapidly activated (to increase serum levels) and then are so rapidly exhausted (such that serum levels decrease) as the vasopressin axis in sepsis.
Ironically, even though vasopressin levels fall to very low levels in septic shock, some evidence suggests that low plasma levels of vasopressin may not be associated with increased mortality. A recent study found that plasma levels of provasopressin were signifi cantly higher in nonsurvivors of septic shock (Table 1) [53]. Th erefore, it is intriguing to speculate that this lack of association of vasopressin levels and mortality could be interpreted to indicate that the decrease of vasopressin levels could be adaptive. Th e studies of levels of vaso pressin and outcome must also be interpreted with an understanding of when vasopressin levels were measured during the course of illness, because changing levels of vasopressin (increased early and decreased later) could be linked to outcomes.
Th e discussion of appropriate levels of vasopressin in septic shock leads logically to the following fundamental question regarding the design of vasopressin intervention studies. Should vasopressin be infused as very-low-dose replacement therapy (opotherapy), at higher doses for vasopressin's vasoconstrictive properties, or at supraphysio logic levels? At low plasma levels of vasopressin (<10 pmol/l), vasopressin's V2 receptor antidiuretic actions predominate. At higher levels, the vasocon strictive proper ties of vasopressin are initiated and become progressively predominant. Th ere is also, however, vaso dilation caused by vasopressin (caused by vaso pressin's stimulation of AVPR1a, AVPR1b, and oxytocin receptors).
Th e early increase in vasopressin release in septic shock is modulated by inducible nitric oxide synthase-induced increase of nitric oxide [54,55]. Heme-oxygenase activation in sepsis attenuates the increase of AVP in septic shock [56].
Copeptin, a 39-amino-acid glycopeptide, is the C terminal of provasopressin (similar to the C peptide of insulin [6]) and is stable in plasma, and is therefore easier to measure than vasopressin. Some studies have therefore suggested that copeptin could be used as a surrogate for measure ment of vasopressin [57][58][59][60]. Landry and colleagues fi rst discovered that vasopressin levels are inappropriately low in septic shock [45,49] (Table 1). Other workers found that vasopressin levels were lower in emergency patients who progressed to septic shock (3.6 pg/ml) compared with patients who had sepsis (10.6 pg/ml) and severe sepsis (21.8 pg/ml) [61].
Critically ill patients have higher copeptin levels than normal volunteers [57,59,60,62,63]. Morgenthaler and colleagues found increased copeptin levels in human sepsis and septic shock (means: normal volunteers, 4 pM; critically ill patients, no sepsis, 27 pM; sepsis patients, 50 pM; severe sepsis patients, 74 pM; septic shock patients, 171 pM) [57]. Furthermore, patients who had hemo dynamic dysfunction had higher vasopressin levels (and copeptin levels) than patients without hemodynamic dysfunction (AVP, 14.1 pg/ml versus 8.7 pg/ml). Patients who are post cardiovascular surgery usually have higher vasopressin and copeptin levels than patients with septic  [58]) and in one study of pediatric septic shock [64]. Lodha and colleagues [64] found that serum vasopressin levels in pediatric septic shock are similar to the results of Jochberger and colleagues [53]. Vasopressin levels were increased in septic shock (116 pg/ml) and severe sepsis (106 pg/ml), and vasopressin levels did not change over 96 hours. Do serum levels of copeptin and vasopressin correlate tightly, such that copeptin is a surrogate for vasopressin? I suggest copeptin levels should not be used as a surrogate for vasopressin levels at this time without further evaluation.
Th ere are fascinating interactions of vasopressin with the baroreceptor refl ex and neural control of the cardiovascular system. It is well known that vasopressin infusion causes bradycardia in normal animals and humans because of stimulation of the baroreceptor refl ex by vasopressin. In addition, infused vasopressin also enters the brain and modulates the sensitivity of the baroreceptor refl ex. Vasopressin is also an important neurotransmitter in the brain that regulates the brainstem control of the cardiovascular system. Vaso pressin is also a key regulator of the behavioral response to stress.

Eff ects of infused vasopressin in septic shock
Vasopressin infusion increases vasopressin levels rapidly [48,49]. Th e eff ects of vasopressin infusion on copeptin levels are not known. Responses to vasopressin and norepinephrine, however, are modifi ed in sepsis. One fundamental concept shown repeatedly in animal studies of sepsis is that the eff ects of vasopressin (both benefi cial and adverse) in models of sepsis depend critically on the volume status of the animal [65]. For example, volume loading in models of sepsis is critical to minimizing and ameliorating the adverse mesenteric vasoconstriction that can produce intestinal ischemia [12]. Th ese observations highlight the central importance of adequate volume resuscitation of patients before infusion of vasopressin is considered.
Fecal peritonitis was used to evaluate the vascular eff ects on mesenteric artery vasoconstriction in response to norepinephrine, vasopressin, and a selective AVPR1a agonist (F180) [12]. Th e vasoconstrictor responses to vasopressin and the AVPR1a agonist were enhanced where as responses to norepinephrine were severely impaired. Th ere thus appeared to be vasopressin hypersensitivity in the mesenteric circulation.
Nakajima and colleagues evaluated vasopressin, norepi nephrine, and l-arginine eff ects on intestinal circulation in a septic shock model [66]. Lipopolysaccharide decreased perfused intestinal villi fl ow; and vasopressors increased the mean arterial pressure (MAP) and prevented further falls in perfused intestinal villi. Knotzer and colleagues studied vasopressin in an acute endotoxin model of septic shock and found decreased jejunal oxygenation [67]. Th ere was no eff ect (that is, no further worsening) from vasopressin.
In contrast, other studies have found decreased vascular responsiveness to vasopressin in models of septic shock [68]. Th ere was a synergistic eff ect of vasopressin and norepinephrine on vascular responsiveness.
Th e eff ects of vasopressin diff ered in several vascular beds in sepsis [69,70]. Renal blood fl ow was lower with norepinephrine than vasopressin; hepatic blood fl ow was lower with vasopressin than norepinephrine, while coronary blood fl ow was similar with vasopressin and norepinephrine.
Patients who have septic shock have decreased forearm vascular resistance, as a component of decreased systemic vascular resistance. Th e forearm vasoconstrictor response to phenylephrine (α-agonist) and angiotensin II was impaired whereas the response to vasopressin was enhanced in septic shock compared with normal [71].
Th e side eff ects of vasopressin studied in animal models of septic shock are confl icting because of diff erences in models (acute, type of septic insult, chronic), doses of vasopressin, and duration of studies (hours versus days). For example, vasopressin infusion (0.06 units/kg/hour) increased the MAP but also decreased mesenteric blood fl ow by one-half and decreased microcirculatory fl ow to gastric mucosa (by 23%) and to jejunal mucosa (by 27%) in porcine fecal peritonitis [79].
Adverse eff ects of vasopressin reported in studies of humans include mesenteric ischemia, myocardial ischemia, digital ischemia, skin necrosis, and hyponatremia, to name a few [62,63,67,76,78,80]. Th ese studies were small, nonblinded, and often uncontrolled, however, so the true risks of low-dose vasopressin infusion in septic shock are unknown. Vasopressin has complex eff ects on the immune and infl ammatory responses. Ironically, commonly used catecholamines (epinephrine and, to a lesser extent, norepinephrine) induce an infl ammatory response in hepatocytes [81] similar to lipopolysaccharide. Eff ects such as these could be harmful directly or indirectly in patients who have septic shock.
Th e eff ects of infused vasopressin on endogenous vasopressin synthesis and release are not known. Early withdrawal of vasopressin infusion in septic shock can lead to a rapid drop in vasopressin levels (Figure 3), sometimes to such low levels of vasopressin that hemodynamic instability could recur. Th e risk of vasopressin defi ciency after withdrawal of vasopressin infusion may be increased in patients who are also on corticosteroids because corticosteroids inhibit vaso pressin secretion.

The Vasopressin and Septic Shock Trial
VASST, a randomized controlled trial of vasopressin versus norepinephrine in patients with septic shock, was performed to address uncertainties regarding vasopressin in septic shock [52]. Th e primary hypothesis was that vasopressin, compared with norepinephrine infusion, would decrease 28-day mortality from 60% to 50%. An important secondary hypothesis was that the eff ects of vasopressin, compared with norepinephrine, would be more pronounced in patients with more severe septic shock. We therefore stratifi ed patients into less and more severe septic shock.
We studied adults with septic shock on norepinephrine treatment. Septic shock was defi ned by infection, two or more of the systemic infl ammatory response syndrome criteria [82], at least one new dysfunction, and vasopressor support (5 μg/minute norepinephrine (or equivalent) for 6 hours). Exclusion criteria are listed in the online supplement [52]. It should be noted that the MAP of patients at baseline in VASST was 72 mmHg, suggesting to some that the patients may not have been in septic shock. All patients were on at least 5 μg/minute norepinephrine (or equivalent), however, which satisfi es the commonly used defi nitions of septic shock [83][84][85][86].
Th e study drugs were vasopressin or norepinephrine; stratifi cation variables were the center and severity of shock. Less severe septic shock was defi ned as infusion of 5 to 14 μg/minute norepinephrine (or equivalent) while more severe septic shock was defi ned as ≥15 μg/minute norepinephrine (or equivalent) in the hour before randomization. Blinded vasopressin infusion was titrated from 0.01 U/minute to 0.03 U/minute, while blinded norepinephrine infusion was titrated from 5 μg/minute to 15  Patients enrolled in VASST were severely ill (for example, high Acute Physiology and Chronic Health Evaluation II scores, high fraction with new organ dysfunction, high norepinephrine infusion rates). Th e MAP was similar in the two treatment groups (Figure 4a). Th e heart rate was signifi cantly lower in the vasopressin group than in the norepinephrine group (fi rst 4 days) ( Figure 4b). As in previous human studies, vasopressin infusion allowed a rapid decrease of the norepinephrine dose (over the fi rst 4 days) ( Figure 5).
Plasma vasopressin was extremely low at baseline (median 3.2 pmol/l) and did not change in the nor epinephrine group. Low-dose vasopressin infusion increased vasopressin levels to 73.6 pmol/l (median at 6 hours) and 98.0 pmol/l (24 hours) (Figure 3). Th ese levels of vasopressin may be supraphysiological because those levels of vasopressin are rarely observed in the very acute phase of septic shock. Some studies found that vasopressin levels increased 200-fold, however, suggesting the circulating levels of vasopressin in VASST were not supraphysiologic.
Vasopressin may have decreased mortality in patients who had less severe septic shock in VASST. At baseline, patients who had less severe septic shock had lower Acute Physiology and Chronic Health Evaluation II score, higher arterial pressure, lower lactate, lower nor epinephrine dose, and a lower proportion presenting organ dysfunction than patients who had more severe septic shock. Th ere were trends suggesting an association of vasopressin infusion with decreased 28-day (and 90-day) mortality in less severe septic shock. Post hoc analyses stratifi ed by indicators of severity of shock suggested a possible advantage of vasopressin. Stratifi cation by lactate quartile and by number of vasopressors at baseline was signi fi cant. In distinct contrast, there were no diff erences in mortality in more severe septic shock between vaso pressin and norepinephrine. Th e test for the interaction of treatment group by severity of shock subgroup, however, was not signifi cant (P = 0.10).
Th e potential benefi ts of vasopressin (compared with norepinephrine) in patients who had less severe septic shock were somewhat delayed so might not have been directly related to rapid reversal of septic shock. Kaplan-Meier curves show that the vasopressin-treated patients separated from the norepinephrine-treated patients at about day 10 (Figure 7). Th is delayed separation of the Kaplan-Meier curves could be due to vasopressininduced delayed benefi t of improved shock, delayed benefi ts of the decreased dose of norepinephrine, or eff ects of vasopressin on infl ammation and immunity. Serious adverse event rates were similar in the vasopressin and norepinephrine groups (10.3% versus 10.5%). Vasopressin infusion may increase the risk of cardiac arrest, yet in VASST there were more cardiac arrests in the norepinephrine group than in the vasopressin group (2.1% versus 0.8%, P = 0.14). Other adverse eff ects of vasopressin and norepinephrine are decreased cardiac output, mesenteric ischemia, hyponatremia (vaso pressin), skin necrosis and digital ischemia. In VASST, there was a trend to more digital ischemia in the vasopressin group versus the norepinephrine group (2.0% versus 0.5%, P = 0.11).
VASST studied the eff ects of vasopressin on patients who were already receiving norepinephrine infusion. Deter mining whether it is the temporal infl uence of the combination of vasopressin plus norepinephrine or of vasopressin alone that was potentially benefi cial in VASST is therefore diffi cult. Studies of vasopressin as the fi rst and only agent versus vasopressin added to -and allowing tapering of -norepinephrine would resolve such controversy.
Some would argue that it may have been unrealistic to design a randomized controlled trial to detect an ARR of Values are mean ± standard deviation (SD). Heart rate was signifi cantly lower in the vasopressin group than in the norepinephrine group over the fi rst 4 days (P < 0.001). There were no statistically signifi cant diff erences between the norepinephrine and vasopressin groups in MAP.  10% in septic shock. Indeed, if the sample size was larger, VASST could have detected a statistically signifi cant ARR of perhaps 5% -which some would argue would be clinically signifi cant and lead to changes in practice. For example, a sample size of 2,286 per group would have been required to have 80% power to detect a statistically signifi cant (P <0.05) decrease in mortality in VASST from 39% (norepinephrine group) to 35% (vasopressin group). Similarly, a sample size of 73 per group would detect a decrease of mortality from 39% to 32%.
Strengths of VASST were the multicenter design, the large sample size powered for mortality, blinding, the use of low-dose vasopressin, blinded evaluation of serious adverse events, well-defi ned inclusion and exclusion criteria, and assessment of pharmacokinetics for several days of vasopressin infusion. Some limitations, however, do apply to VASST. First, the vasopressin infusion rate was predetermined for a low dose. Second, we could not measure plasma vasopressin levels to guide the dose or duration of drug infusion. Th ird, the MAP at baseline in VASST was about 72 to 73 mmHg, so VASST evaluated low-dose vaso pressin as a catecholamine-sparing drug. Finally, the mean time from meeting the VASST inclusion criteria to study drug infusion was 12 hours -Rivers and colleagues identifi ed benefi ts of early goal-directed therapy when applied in the fi rst 6 hours [87], so delays greater than these 6 hours may be important. Th ere is some evidence from VASST that early treatment with vasopressin may optimize its benefi ts. Th ere was a trend to decreased mortality with vasopressin (33.2% versus norepinephrine 40.5%, P = 0.12) in patients treated within 12 hours. Early vasopressin infusion may optimize benefi ts similar to early goal-directed therapy [87].

Vasopressin as the initial agent in septic shock
Some would ask whether vasopressin infusion could be used as the first agent in patients who have not responded to adequate volume resuscitation, especially since vasopressin was most effective in patients who had less severe septic shock in VASST. VASST did not evaluate vasopressin as the initial agent in septic shock.
A retrospective cohort study of patients who had septic shock who received either norepinephrine (n = 49), vaso pressin (n = 50), or dopamine (n = 51) as the fi rst vaso active agent in septic shock found no diff erence in mortality (65%, 52%, and 60%, respectively) [88]. Further more, there are no studies of the safety of vasopressin as the fi rst agent in septic shock. We therefore do not believe there is yet adequate evidence to recommend vasopressin as the initial agent in septic shock.

Interactions of vasopressin infusion and corticosteroid treatment in septic shock
Th ere have been remarkably few large studies of interactions of the commonly used therapies in critical care and the eff ects on mortality. Many of the patients in VASST were treated with corticosteroids because patients had septic shock. We therefore evaluated the interaction of vasopressin infusion, corticosteroid treatment, and mortality of septic shock in VASST [89]. Low-dose vasopressin infusion plus corticosteroids signifi cantly decreased 28-day mortality compared with corticosteroid plus norepinephrine treatment (44.7 to 35.9%, P = 0.03; P = 0.008 interaction statistic). In contrast, if patients were not treated with corticosteroid, then vasopressin may have increased mortality compared with norepinephrine (33.7 versus 21.3% respectively, P = 0.06). In patients who received vasopressin infusion, corticosteroids increased plasma vasopressin levels signifi cantly to 33% at 6 hours (P = 0.001) and to 67% at 24 hours (P = 0.032) versus patients who did not receive cortico steroids.
To the knowledge of the VASST investigators, this substudy of VASST is the fi rst clinical study of interactions of vasopressin and corticosteroids on mortality and organ dysfunction in human septic shock. Many patients probably receive the combination of vasopressin and corticosteroids in clinical practice. Despite this, there are few studies of the interaction of vasopressin and corticosteroids in septic shock. Th e eff ects of corticosteroids on vasopressin levels and effi cacy are controversial. Studies fi nd that corticosteroids increase vasopressin mRNA and restore responsiveness to vasopressin, have no eff ect on vasopressin levels, and even suppress vasopressin gene expression [90][91][92][93].
Th e eff ects of vasopressin on the corticosteroid axis are clearer [90,94,95]. When vasopressin binds to AVR1b (V3 receptor), vasopressin increases corticotroph responsiveness to corticotropin-releasing factor, thus increasing ACTH even in conditions of stress when corticosteroid levels are increased [73,75,90,94,95]. Th e vasopressininduced increase of ACTH (unlike eff ects of cortico tropinreleasing factor on ACTH) is resistant to cortico steroidnegative feedback because binding to AVR1b is coupled to phospholipase C and is not regu lated by corticosteroid levels. Vasopressin infusion did not change serum ACTH or cortisol levels in a cohort of septic shock [73,75].
It is not known how vasopressin plus corticosteroids (versus norepinephrine plus corticosteroids) could have altered mortality rates. Potential mechanisms could be that corticosteroids increased vasopressin levels or enhanced responsiveness to vasopressin [91], or the combination of vasopressin and corticosteroids altered infl amma tion and immunity benefi cially. Annane suggested that inter actions with nitric oxide could explain this inter action [96]. Furthermore, vasopressinases such as LNPEP catalyze the metabolism of vasopressin in the circulation. Interestingly, LNPEP and other vaso pressinases are corticosteroid responsive [96]. Th e eff ects of corticosteroids on vasopressin receptors have not been extensively studied. Dexamethasone attenuates the downregulation of AVPR1a receptors in sepsis by inhibition of NFκB.

Use of higher doses of vasopressin
Several investigators have explored the use of higher doses of vasopressin in septic shock [62,63,[73][74][75], and indeed have argued that this is one reason why vasopressin might not have decreased mortality in VASST [97]. Cohort studies, case-control studies, and small randomized controlled trials have used doses of vasopressin of about 0.06 units/minute (compared with 0.03 units/minute in VASST), and found that vasopressin signifi cantly decreased use of norepinephrine and increased blood pressure [62,63,73,75]. Some of these studies, however, also detected potentially adverse side eff ects of vasopressin such as decreased intestinal mucosal perfusion [98,99], increased bilirubin [62,63], increased serum transaminases [62,63,73,75], and decreased platelet counts [62,73]. Note that some of these studies used much higher doses than recommended in human septic shock [98][99][100]. We have agreed with the suggestion [97] that a large randomized trial of these higher doses of vasopressin (0.06 units/minute) to address safety and effi cacy questions in septic shock is meritorious [97].
Like vasopressin [73], terlipressin may decrease cardiac output in septic shock [101], especially if there is myocardial depression or pre-existing myocardial dysfunc tion (such as pre-existing congestive heart failure, ongoing myocardial ischemia) [126]. Th e decrease of cardiac output due to terlipressin is prevented in animals [126] and humans by co-administration of dobutamine [111], which increases the left ventricular stroke work index [111]. Furthermore, terlipressin may increase coronary vasoconstriction [127] and induce myocardial ischemia because both vasopressin and terlipressin may potentiate adrenergic vasoconstriction in coronary arteries [128] (as well as mesenteric arteries [129]). Terlipressin may also have direct coronary vasoconstrictive properties because of binding to AVPR1a [127]. Terlipressin has been associated with ventricular arrhythmias [130], with decreased platelet counts [101,122] andsimilar to vasopressin [66,79,107] -with mesenteric ischemia, in part because both agents may potentiate adrenergic vasoconstriction in the mesenteric circulation [129].

Conundrums and one view to consensus
Th ere are conundrums in the vasopressin literature.
Large clinical trials such as VASST suggest that vasopressin may be more eff ective in patients who have less severe septic shock and in combination with corticosteroids. Yet experimental studies suggest that AVPR1a sensitivity and the prevalence of adrenal dysfunction increase with severity of shock. Th e confl icting evidence about AVPR1a sensitivity (some studies suggest increased AVPR1a sensitivity while others found less sensitivity in models of sepsis) suggests this is an ongoing controversy. Regarding the interaction of vasopressin and corticosteroids, perhaps the benefi cial interaction is not related to the presence of adrenal dysfunction but rather is related to corticosteroid-induced increase of vasopressin levels [52,89]. Vasopressin levels were higher in patients who had preserved adrenal function -and higher corticosteroid levels -compared with patients who had impaired adrenal function (Table 1).
Another puzzle is the observation that the effi cacy of corticosteroids is only moderate in moderate septic shock (CORTICUS) [86] and is greater in more severe septic shock. We have recently shown that corticosteroids appear to increase plasma vasopressin levels in both less severe and more severe septic shock, but the eff ect was statistically signifi cant only in less severe septic shock. Th e benefi cial interaction of vasopressin infusion and corticosteroid treatment may therefore span the spectrum of severity of shock.
Another question is the relationship of vasopressin to dopamine, to dobutamine plus norepinephrine, and to epinephrine -considering that VASST focused on patients of whom the majority were already on nor epinephrine infusion. In VASST, some patients were on dopamine or epinephrine instead of norepinephrine; however, the numbers of these patients were too small to assess accurately the eff ects of vasopressin on patients who were receiving these vasoactive agents. It is likely that the fi ndings of studies of vasopressin eff ects on norepinephrine dose also apply to dopamine when dopamine is used as a vasopressor (that is, doses >5 μg/ kg/minute). It is quite likely that vasopressin can be used eff ectively with dobutamine because several animal and human studies have shown that the combination of vasopressin and dobutamine is benefi cial in limiting any decrease of cardiac output, oxygen delivery, and mixed venous oxygen saturation caused by vasopressin. It is also likely that vasopressin may also be used in patients who are already on epinephrine infusion. Th e combination of dopexamine and vasopressin attenuated decreases of the cardiac index and oxygen delivery -compared with vasopressin alone -but dopexamine also decreased blood pressure compared with vasopressin; this suggests that dopexamine may not be a good agent to attenuate cardiovascular adverse eff ects of vasopressin.
Clearly there is a common, profound defi ciency of vasopressin in septic shock. However, it is not known when and how quickly this vasopressin defi ciency recovers. Siami and colleagues addressed this important question by measuring plasma vasopressin levels of normonatremic patients who had septic shock for more than 72 hours who were given a stimulus: osmotic stimulation using 500 ml hypertonic saline [131]. Interestingly, 52% (17 of 33 patients) of patients were nonresponders to this stimulation. In responders, plasma vasopressin increased from 4.8 to 14.4 pg/ml, a rate of increase of 6.2 pg/ml/hour. In contrast, nonresponders' plasma vasopressin did not increase signifi cantly (from 2.8 to 4 pg/ml, a rate of increase of 0.7 pg/ml/hour). Nonresponders were more frequently bacteremic and had hepatic dysfunction compared with responders. Hypertonic saline lead to comparable changes in central venous pressure, MAP and systolic arterial pressure in responders and nonresponders; thus, it is unlikely that the diff erences in response of plasma vasopressin levels were due to diff erences in baroreceptor refl ex stimulation. Osmoregulation by vasopressin is therefore impaired in about one-half of this small sample of patients who had sustained septic shock. It is likely that vasopressin regulation of blood pressure is similarly impaired for at least this duration.
What is the status of the vasopressin axis shortly after patients are weaned off vasopressin infusion? Bauer and colleagues found that patients dropped their blood pressure more often when vasopressin infusion was stopped before stopping infusion of norepinephrine [132]. Th eir study suggests there can be ongoing inadequate vasopressin axis reserve. Th is hypothesis is not directly known, but the vasopressin axis is probably only partially recovered for some time after discontinuation of vasopressin infusion and a secondary insult (for example, new nosocomial infection) might lead to septic shock more readily because of continued impaired vasopressin axis responses.

Newer alternative strategies to support ventricular function in patients whose ventricular function declines on vasopressin
Vasopressin infusion has been reported to decrease cardiac output in some patients who have septic shock. Dobutamine infusion is often added to increase cardiac output [133]. Levosimendan is a novel inotropic agent that acts by sensitizing calcium channels leading to vasodilatation, and often (but not always) leading to increases in cardiac output. Rehberg and colleagues showed that levosimendan plus vasopressin improved aspects of hemodynamic para meters (such as the left ventricular stroke work index) and net fl uid balance in sheep that had fecal peritonitis compared with vasopressin therapy alone [134]. Th e combination of vasopressin plus levosimen dan also increased global perfusion (as assessed by mixed venous oxygen satura tion) and decreased pulmonary artery pressure. Th is novel combination also appeared to improve the net fl uid balance and prevented decreases in colloid osmotic pressure compared with vasopressin alone. Levosimen dan is not yet available for clinical use in North America, but its study is intriguing and bears further clinical evaluation.
If patients suff er a decrease of cardiac output and tissue perfusion in response to vasopressin, therefore, one should re-evaluate the volume status -volume resuscitate if defi cient, and if volume replete the current evidence supports adding dobutamine infusion to augment cardiac output.

Genomics and vasopressin
Recent studies suggest that there could be notable genomic variation in vasopressinases (also known as LNPEP), the enzyme that metabolizes vasopressin (as well as oxytocin and GLUT4). Furthermore, Nakada and colleagues found that patients who had the genotype TT of the single nucleotide polymorphism rs4869317 of the LNPEP gene had increased mortality (compared with patients who had the AT/AA genotypes) in a discovery cohort of patients who had septic shock, and that this fi nding was replicated in an independent cohort of patients who had septic shock [135]. Th is marker (rs4860317 TT) was also associated with increased cardio vascular dysfunction. Furthermore, these authors discovered that patients who were on a vasopressin infusion who had the rs4869317 TT genotype had lower plasma vasopressin concentrations because of increased vasopressin clearance [135]. To summarize, the TT genotype of rs4869317 was associated with increased mortality and increased vasopressin clearance in patients who had septic shock.
Patients who have the TT genotype of rs4869317 would also be expected to have altered AVPR1a and AVPR2 responses to vasopressin because of altered vasopressin levels available to these diff erent vasopressin receptors. Indeed, the plasma sodium concentration was higher in the fi rst 48 hours of cardiovascular surgery in patients who had the rs4869317 TT genotype compared with patients who had the AT/AA genotypes. Th is fi nding aligns with the diminished vasopressin levels and increased mortality rates of patients with the TT genotype of rs4869317 who had septic shock. Th e genomic variation of the LNPEP gene of patients could therefore alter the response to vasopressin (and other vasopressin analogues) in septic shock, postcardiovascular surgery and other conditions such as hepatorenal syndrome.
Th e role for the use of genetic testing to provide more patient-centered care is exciting, requires more investigation and could be of benefi t in the selection of patients for treatment with vasopressin or vasopressin analogs.

New conditions in which there is a vasopressin defi ciency
Patients with end-stage liver disease characteristically have a vasodilated circulation that can mimic sepsis and septic shock. Th erefore it is reasonable to hypothesize that there could be a vasopressin defi ciency in such patients. Wagener and colleagues evaluated a cohort of such patients by measuring plasma vasopressin levels and responses to infusion of vasopressin [136]. Plasma vasopressin levels were inappropriately low in patients who had end-stage hepatic disease undergoing liver transplantation compared with control surgical patients. Vasopressin increased systemic vascular resistance and arterial pressure. Patients who have end-stage hepatic disease thus appear to have vasopressin defi ciency and may benefi t from vasopressin infusion (once fl uid resuscitation is adequate).
Th ese observations of potential benefi ts of vasopressin in end-stage liver disease are congruent with the studies showing benefi ts of terlipressin in patients who have severe liver disease and renal dysfunction -hepatorenal syndrome.

More selective V1a agonism in septic shock: another approach
As discussed, there could be advantages to more selective V1a agonism treatment in septic shock such as purer vasoconstriction, limited antidiuretic eff ects, lack of stimulation of thrombosis (because of lower increase of von Willebrand multimers) and perhaps better protection from increased permeability. Rehberg and colleagues designed a clever study in which they gave a V2 receptor antagonist (propionyl-1-d-(Tyr(Et)2-Val4-Abu6-Arg8,9)vasopressin) with or without a V1a agonist (arginine vaso pressin) or placebo to chronically instrumented sheep who had early septic shock [137]. Th is V2 receptor antagonist with arginine vasopressin maintained hemodynamics (cardiac index and MAP) and increased the left ventricular stroke work index. Most notably, the combination of the V2 receptor antagonist with arginine vasopressin (com pared with arginine vasopressin alone) improved markers of tissue perfusion (for example, pH and arterial lactate), while also attenuating liver and renal dys function. Markers of lung injury (hemeoxygenase and 3-nitro tyro sine concentrations) were also more favorably protected in the animals that received this combination of V2 receptor antagonist with arginine vasopressin compared with arginine vasopressin alone. Also intriguing is the observation that the administration of arginine vasopressin alone increased plasma vasopressin levels, whereas the combi nation of this V2 receptor antagonist with arginine vasopressin increased plasma vasopressin levels to a lesser degree.
Th is innovative animal model study provides further evidence to support further clinical investigation into the role of selective V1a agonism in human septic shock.

Clinical perspectives
What should the clinician do in considering use of vasopressin in patients who have septic shock? My current recommendations are to consider use of vasopressin in patients who have less severe septic shock because of the encouraging eff ects of vasopressin in these patients in VASST [52].
Should vasopressin levels be measured in patients who have septic shock? One obstacle is that vasopressin can only be measured by radioimmunoassay, which is relatively cumbersome and has too great a turnaround time for patients who have septic shock. I would therefore not recommend measurement of vasopressin levels (or copeptin levels, as already discussed on page 6 above) at this time for clinical purposes in patients who have septic shock. I would recommend that vasopressin be given as a continuous infusion not exceeding 0.03 units/minute, because that was the dose used in VASST.
Should vasopressin be used as fi rst-line therapy in hypotensive septic patients? I believe this is a useful question for clinical research but not yet for clinical practice.
Should vasopressin and corticosteroids be given together? Th e potentially benefi cial interaction of vaso pressin infusion and corticosteroid treatment on mortality was discovered in a post hoc substudy of VASST and could be a false positive, and so is hypothesisgenerating at this time. Th e Surviving Sepsis Guidelines [85] state that norepinephrine and vasopressin are similar and that vasopressin should be considered in patients who have septic shock.

Future research
Some themes for future research could include assessment of vasopressin as the fi rst-line agent in septic shock, evaluation of the renal eff ects of vasopressin, further assessment of adverse eff ects of vasopressin versus norepinephrine, and use of the more specifi c AVPR1a agonist. Preliminary reports suggest that an AVPR1a agonist causes less third-spacing of fl uid (vascular leak syndrome), could reduce release of the procoagulant protein von Willebrand factor (released by stimulation of AVPR2), and decreases mortality compared with vasopressin in two diff erent models of septic shock.

Conclusions
Vasopressin stimulates a family of receptors (AVPR1a, AVPR1b, and AVPR2). AVPR1a is responsible for the vasoconstriction associated with vasopressin. Sepsis may downregulate all three vasopressin receptors. Vasopressin levels increase early in septic shock because hypotension is the most potent stimulus for release of stored vasopressin. Vasopressin levels then decline rapidly -due to depletion of stored vasopressin.
Responses to vasopressin and norepinephrine are modifi ed in sepsis and, furthermore, the eff ects of vasopressin diff er in diff erent vascular beds in sepsis. Vasopressin infusion increases blood pressure, decreases norepinephrine dose requirements, and improves renal function. Previous clinical studies of vasopressin in septic shock, however, were small or not controlled. Adverse eff ects of vasopressin include mesenteric ischemia, myocardial ischemia, digital ischemia, skin necrosis, and hyponatremia.
VASST, a randomized controlled trial of vasopressin versus norepinephrine in septic shock, was performed to address uncertainties regarding effi cacy and safety of vasopressin in septic shock. Th ere was no signifi cant diff erence between vasopressin-treated and norepi nephrine-treated groups in 28-day mortality (35.4% versus 39.3%, P = 0.26; ARR = -2.9 to 10.7%) or in organ dysfunc tion rates. Vasopressin may have decreased mortality in patients who had less severe septic shock in VASST. Serious adverse event rates were similar in the vasopressin and norepinephrine groups.
Vasopressin therefore appears equally as safe as norepinephrine in septic shock, and may improve mortality in less severe septic shock. Th ere have been remarkably few large studies of interactions of the commonly used therapies in critical care and the eff ects on mortality. Low-dose vasopressin infusion plus corticosteroids signifi cantly decreased 28-day mortality compared with corticosteroids plus norepinephrine (44.7 to 35.9%, P = 0.03; P = 0.008 interaction statistic). Th is interaction is potentially important and requires further study.

Summary
Low-dose vasopressin may be eff ective in patients already receiving norepinephrine who have less severe septic shock -such as patients with modest norepinephrine infusion (5 to 15 μg/minute) or low serum lactate levels. Additional studies such as large randomized controlled trials are necessary to address questions regarding higher doses of vasopressin, use of vasopressin as the fi rst agent in septic shock, use of terlipressin and other more specifi c AVPR1a agonists, and the interaction of vasopressin and corticosteroids in septic shock.