During the study period, 249 patients had CS inclusion criteria. Population description and its management are summarized in Table 1. Briefly, most patients were intubated (89%), and received catecholamines (97%). In-hos- pital mortality was 46%, as expected in CS patients with such high severity markers, including SAPS II at 64.

Hypocalcaemia was present in 117 (47%) of the patients during ICU course. Hypercalcaemia was noticed in 13 (5%) of the patients upon admission, was iatrogenic in 25 (10%) additional patients and was never severe. Lower blood calcium level during ICU stay was associated with increased length of stay (r² = 0.01, p = 0.003), extra- renal replacement therapy (ERRT), mechanical ventilation duration (r² = 0.01, p < 0.001), SAPS II (p = 0.03). There was a strong trend for lower blood calcium level in non-survivors than in survivors (p = 0.053, Table 1). However, in a multivariate analysis the association was not significant. Calcium blood level has been normalized in 84 (72%) of the 117 patients with hypocalcaemia, in 24 hours (8 - 72) on average. Ischemic events or troponin release were similar in all groups (hypercalcaemia, hypocalcaemia, corrected or not). The correction of the hypocalcaemia, the time of correction and hypocalcaemia were not associated with any improvement in hemodynamic parameters, mechanical ventilation duration, or length of stay. Only arrhythmias were associated with hypocalcaemia duration before normalization.

Intra-aortic balloon pump (IABP) was inserted in 76 patients (31%) and was not statistically associated with in- hospital mortality (OR 0.67; CI 0.4 - 1.2; p = 0.2). Median duration of IABP support was 3 days (2 - 6). Extra-corporeal life support (ECLS) was used in 20 patients for a median 8 days (3 - 12). ECLS was always used for reversible cause of shock. Nine of the ECLS treated patients died (45%) and 11 survived. Six ECLS patient had PH or right ventricular (RV) failure: 2 died (33%) and 4 survived. The survivors had RV AMI, massive mitral regurgitation, fulminant myocarditis or pheochromocytoma while the non-survivors had mitral valve replacement or pulmonary embolism. Of the 18 patients with RV failure, 12 died (67%, in whom 4 of the 6 pulmonary embolism, 67%) and 6 survived. Though, mortality of patients with RV failure related CS was high (10/12, 83%) but not statistically different from patients treated with ECLS (Fischer, p = 0.1). The use of inhaled nitric oxide in 17 patients was not associated with a better outcome (OR 0.73; CI 0.3 - 2; p = 0.5).

Most of patients received catecholamines (241 patients, 97%): single (N = 167), association of two (N = 57), three (N = 12), or four (N = 5) vasoactive agents. When single vasoactive agent was used, it was an inotrope including epinephrine (N = 115), dobutamine (N = 33), or dopamine (N = 1), but also a predominant vasopressor, norepinephrine (N = 18). More common associations were epinephrine + dobutamine (N = 35), norepinephrine + dobutamine (N = 14). The choice of initial catecholamines is summarized in Table 1.

The prognosis analysis is shown in Table 1. In the multivariate analyses, SAPS II (OR 1.037; CI 1.013 - 1.056; p = 0.0001), PH (OR 4.8; CI 1.3 - 17; p = 0.02), ERRT (OR 2.9; CI 1.3 - 6; p = 0.006), and dobutamine infusion (OR 0.44; CI 0.2 - 0.96; p = 0.04) were significantly associated with in-hospital mortality.

In patients surviving the acute phase, prescription of betablockers (p = 0.002), ACEI (p = 0.0001), and statins (p = 0.01) was associated with improved outcome and decreased in-hospital mortality.

In this cohort of CS patients, when used as the first choice catecholamine, dobutamine is significantly and independently associated with an improved outcome and a doubled hospital survival. This beneficial association was present whatever the etiology of CS. In order to increase cardiac output, after volume correction, patients received dobutamine or epinephrine―or the combination of both―as other inotropes like dopamine or phosphodiesterase-3 inhibitors were exceptionally prescribed in our ICU [5] . Accordingly, inodilators (namely, dobutamine, levosimendan, and phosphodiesterase-3 inhibitors) have been associated with decreased mortality (OR 0.66; CI 0.55 - 0.80) after propensity score analysis in 988 patients with CS [17] . Levosimendan is still not approved in our country and its use requires a specific authorization; thereby, it was not available in our hospital. Norepinephrine was sometimes added to dobutamine, or used alone. This was the main predominant vasopressor in our study, as we never infused phenylephrine or vasopressin to our patients (one association with terlipressin). Association of catecholamines is often preferred to separate inotrope from vasopressor effect rather than increasing the amount of a single catecholamine [4] . Though, in order to improve contractility, dobutamine (with the association of norepinephrine and/or another vasopressor) might be preferred over epinephrine in the management of patients with CS.

Nevertheless, despite many described side effects like hyperglycaemia, lactic acidosis [4] [8] [18] , or coronary artery thrombosis stimulation [19] , to our knowledge, no difference in major clinical outcome between catecholamines has been published in the CS setting, except increased arrhythmias and mortality with dopamine [3] [20] . In our cohort of patient, dobutamine infusion seems a safe and efficient inotrope agent. It can be the first line inotrope even if as epinephrine, dobutamine infusion leads to platelets activation [21] .

Discharge treatment is of paramount importance. At recovery from shock, beta-blockers should be introduced. Their use after acute severe HF is associated with better outcome [22] . ACEI have hormonal beneficial properties but induce pronounced vasodilation. Resuming ACEI before surgery can decrease hypotensive episodes and catecholamines support. During CS, systemic vascular resistances are sometimes altered and ACEI might be detrimental. However to our knowledge, no clinical study has addressed this issue and rapid re-introduction, after the acute phase, should be considered with caution. Statins-associated outcome improvement during CS has already been reported [23] .

We have also already reported our experiences on phosphorus and coronary stent [24] [25] . Similarly, calcium management in this setting is associated with conflicting results in the literature. Although calcium supplementation can dramatically improve inotropy in some patients, it is not associated with improved outcome in the ICU [26] . Moreover, it has been associated with increased in long-term risk of myocardial infarction. Our data suggest that calcium supplementation does not improve dramatically the evolution and does not lead to ischemic injury in patients with CS. However, even if single electrolytes management could hardly modify patients’ evolution, medical screening for severe electrolytes deficiencies and diligent adequate correction may improve outcome [24] .

Therapeutic hypothermia is regarded as an attractive procedure in post-AMI CS and in order to decrease global oxygen consumption [27] . However, these data deserve scrutiny, as lower initial temperature was associated with increased mortality in our study. This association was found in patients presenting with different causes of shock, i.e. post-myocardial infarction, post cardiac arrest CS. Although for our patients, lower admission temperature was not significantly associated with mortality in the multivariate analysis. Many confounding factors could have impact our results as lower temperatures were recorded in clinically more severe patients. However, we did find an association between a lower admission temperature and the frequency of arrhythmias (p = 0.005), especially ventricular ones (p = 0.04), but not with other complications such as infection or hemorrhage. In our study only few patients reached hypothermic temperatures (<34˚C), we cannot draw firm conclusion but incite to careful evaluation of systematic lowering of body temperature during CS [28] .

In our study, mechanical treatment had no effect on the outcome. IABP has been widely used for mechanical hemodynamic support, it improves coronary perfusion and decrease afterload with a beneficial effect on oxygen consumption. However its effect is still debated with conflicting results on early and late mortality. Recently, in 600 patients with MI and CS, randomized to IABP or no mechanical treatment, IABP proved to have no effect on short term and long term mortality [29] . Our data are similar. In our hospital, ECLS is used in refractory CS in order to correct the hemodynamic disorders before multiorgan failure develops. We don’t have specific protocol and implantation is upon the clinical empirical judgment of the cardiology team in charge (cardiologist, intensivist and cardiac surgeon). ECLS has been implemented in 20 patients. Even if these assisted patients were more severe, mortality was identical (45%), and one might hypothesize that ECLS can be a lifesaving procedure. To prove the ECLS effect on mortality, studies are needed and require patient selection with a precise and widely acceptable definition of refractory CS (as it does not exist, to our knowledge), and careful, precise and identical medical management.

Among the most severe cases of our cohort, non-responders to dobutamine and/or patients with PH may benefit from more aggressive treatment including assist device. Our data suggest that this population may be selected as refractory CS. CS due to isolated RV failure might gain special benefit from peripheral ECLS as it is a good RV unloading system and increasing the post-charge of a “normal” left ventricle might not lead to further insult. Our six patients with isolated RV failure treated with ECLS had a low mortality (33%). It was not statistically significant and need further evaluation but in case of reversible cause of shock (mainly pulmonary embolism, pending recovery from fibrinolysis or embolectomy), ECLS might be considered as a rescue and efficient procedure [30] .