Microcirculatory dysfunction in sepsis is crucial in the pathogenesis of organ dysfunction and consists of a cascade of mechanisms, which involves many cells, such as endothelial cells, smooth muscle cells, red blood cells and leukocytes [1]. New microcirculatory imaging techniques, such as orthogonal polarization spectral (OPS) imaging [2] and its technical successor sidestream dark field (SDF) imaging [3], have allowed direct observation of the microcirculation at the bedside. De Backer et al. [4] were the first to report that the observed microcirculatory alterations in sepsis were associated with morbidity and mortality, irrespective of correction of global hemodynamic variables [5].

Therefore, other strategies that aim to attenuate microcirculatory dysfunction directly at the level of the microcirculation, have been tested. According to Poiseuille's Law, blood flow through a vessel is directly proportional to the driving pressure along the vessel and its radius to the fourth power, and inversely proportional to the length of the vessel and the dynamic blood viscosity. Theoretically, vasodilators should be able to increase blood flow via a change in vessel diameter at the entrance of the microcirculation and recruit microcirculatory perfusion in volume resuscitated patients [6]. Stimulation of endothelium-dependent vasodilation by topical application of acetylcholine was effective in the recruitment of shut-down capillaries in septic patients, thus challenging the concept of vasoplegia [7]. It suggested that endothelial vasodilatory response is intact in sepsis and microvascular thrombosis is not the predominant factor in the observed decrease in microcirculatory blood flow. Thus, an ideal agent to recruit the microcirculation in sepsis should have endothelium modulator and vasodilator characteristics, such as a nitric oxide (NO) donor [8, 9]. However, a recent randomized clinical trial could not confirm the previous observation, that nitroglycerin improved microvascular perfusion in an open label setting [10].

Studies with magnesium (Mg) and magnesium sulphate (MgS) have shown a peripheral (predominantly arteriolar) vasodilator effect with preserved cardiac function, within a large safety margin [11–13], not only indirectly by an endothelium-dependent release of NO [14, 15] but also directly via its ability to induce endothelium-independent vasodilation by a direct action on vascular smooth muscle as a calcium competitor. In addition to vasodilatory effects, infusion of MgS is also associated with other potential pro-microcirculatory effects, such as an increase in red blood cell deformability, reduction of platelet aggregation, anti-inflammatory effects and maintenance of endothelial integrity [16–19].

However, up to date the effect of MgS infusion on microvascular perfusion in septic patients, is unknown. Our study aims to test the hypothesis that MgS infusion may improve sublingual microcirculatory perfusion in patients with severe sepsis and septic shock.

To our knowledge, this is the first study in which the effect of Mg on the human microcirculation is visualized directly in patients with sepsis. The main finding of our study is that MgS infusion had no effect on both sublingual microvascular perfusion and systemic hemodynamic variables in this setting. This is not in line with previous observations, that evaluated the effect of MgS on microcirculatory perfusion under non-septic conditions. In vivo plethysmography of the forearm in healthy volunteers during intra-arterial MgS infusion revealed a marked increased in blood flow. In an animal model with direct in-vivo microscopy of the microcirculation, both topical and intravenous application of Mg-compounds demonstrated vasodilatory effects in mesenteric arterioles (10-20 μm) and venules (15-30 μm), that could not be blocked by cyclo-oxygenase inhibition [24]. This effect was also observed during catecholamine-induced vasoconstriction [25].

The discrepancy between these results and our observations may in part be explained by a difference in technique. Plethysmography incorporates artiolar, capillary and venular blood flow. In sepsis however, a heterogeneous distribution of altered capillary blood flow is observed, often despite maintenance of both artial and venular blood flow. In contrast to plethysmography, SDF enables discrimination between these different vascular compartments. Furthermore, sublingual SDF imaging may not necessarily reflect microvascular perfusion abnormalities in vascular beds of other organs [26]. Whether the sensitivity of software-supported analysis of TVD of vessels equals to direct diameter measurements of vessels during in-vivo microscopy remains to be established and is very sensitive to adequate focus.

The classical response to overcome microvascluar perfusion abnormalities in shock has always been to raise arterial pressure by means of vasopressors, in order to increase the net perfusion pressure over the microvascular compartment. However, according to physiological theory, the drop in perfusion pressure prior to the microcirculation is equal to the increase in artiolar resistance, thus reducing its pro-microcirculatory flow effect [27]. Recent studies indeed failed to demonstrate a pro-microcirculatory effect after a stepwise increase of blood pressure with norepinephrine [28, 29]. Although the alternative concept of vasodilation to recruit microcirculatory perfusion in sepsis is tempting [30], up till now no clinically available vasodilators with pro-microcirculatory characteristics have been discovered. Despite promising results in a small group of patients [9], nitroglycerin showed no effect on microcirculatory perfusion in a randomized placebo controlled trial in patients with severe sepsis and septic shock [10]. Both the latter study and the presented trial have in common that a vasodilator was infused on top of a strict resuscitation protocol, making hypovolemia unlikely. Even under well-descript vasodilatory effects of epidural anesthesia, hypervolemic hemodilution completely blunted microcirculatory changes in flow and capillary density of the vaginal mucosa [31]. The absence of effect of a vasodilator in this setting, both on macro- and microcirculatory variables, may be in line with maximal pre-capillary smooth muscle relaxation prior to the start of the intervention. This is particularly of note since, unlike nitroglycerin, MgS has an additional non-endothelium dependent pathway to exert its vasodilatory capacities, by a direct action on vascular smooth muscle relaxation as a calcium antagonist [15].

Our study has a number of limitations. Due to the open label setting and relatively small numbers a modest effect of MgS infusion on microcirculatory perfusion may have remained unnoticed. Baseline MFI was decreased despite resuscitation, but only to a moderate extend. Although this may have blunted a potential pro-microcirculatory effect of MgS, we only observed a non-significant and weak correlation between baseline MFI and change in MFI in response to therapy. It must be stated however, that post-hoc linear regression analysis itself is limited by the narrow range of variables. We deliberately chose this strategy, since we consider it key not to replace therapeutic macro-hemodynamic strategies by micro-hemodynamic strategies, but to add micro-hemodynamic strategies in case of remaining microcirculatory abnormalities, despite optimal macro-hemodynamic resuscitation. In our study we used a continuous fixed dose of MgS without bolus due to safety reason, since some data show, that higher doses of MgS may inhibit catecholamine release [32, 33]. Endothelium-independent vasodilation is associated with venous Mg concentrations, but the endothelium-mediated vasodilatory pathway of Mg appears to be activated, irrespective of its plasma levels [14]. In the present study there was no correlation between baseline Mg concentrations and the change in MFI pre- and post MgS infusion. We failed to report plasma Mg concentrations after infusion. The observation period after the intervention was without delay. This may have been too short to allow microcirculatory alterations to respond to the intervention. On the other hand, longer observation periods in an open label setting carry a considerable risk of improvement over time, irrespective of the intervention itself. Therefore, we cannot rule out that a higher dose, a different loading regiment, suppletion to a pre-defined plasma Mg concentrations or a longer post-intervention observation period may have revealed more potential of MgS infusion for the improvement of microcirculatory perfusion in sepsis.

In the setting of severe sepsis and septic shock sublingual microcirculatory alterations were observed despite fulfillment of sepsis resuscitation guidelines. After infusion of a limited and fixed dose of MgS, microcirculatory perfusion did not improve over time.