Sudden death remains a major public health issue, despite improvements in prehospital management and standardization of advanced life support through wide diffusion of international guidelines 1 . Both incidence and poor prognosis are striking: according to official statistics, approximately 100,000 people are supported for out-of-hospital cardiac arrest (OHCA) in the United States each year. However, it is estimated that the real number of sudden death is two to three times higher. Even more problematic, less than 10% of patients admitted to the hospital after successfully resuscitated OHCA will leave the hospital without major neurological impairments.

For patients who survive the initial phase of prehospital care, the course is usually marked by two types of events:

1. Syndrome originally described as an early reperfusion syndrome (or "postresuscitation disease"), which usually appears between the 4th and 24th hour in the form of a stereotypical feature whose extreme form involves a state of shock, high fever, and severe biological disorders 2 .

2. Poor neurological prognosis because two thirds of patients who survive the early phase will subsequently develop neurofunctional sequellae, which sometimes progress toward a postanoxic vegetative state and delayed death 3 .

The frequency and intensity of these complications depend largely on the delay of initial treatment, the efficiency of resuscitation, and the time elapsed between collapse and return of spontaneous circulation (ROSC). At that time, CA and its resuscitation is the clinical situation closest to the phenomenon of "ischemia-reperfusion," well known from experimental models 4 . Moreover, this is the only medical situation that enables measurement of clinical acute consequences of global ischemia, targeting simultaneously all tissues and organs. The pathophysiology of this syndrome explains the observed features and justifies the therapeutic interventions that must be performed to achieve a favorable neurological evolution. Optimizing postresuscitation care is of paramount importance, because it represents the last link of the survival chain.

The pathophysiology of postcardiac arrest syndrome is complex and remains partially understood (Figure 1). It seems, however, dominated by a global ischemia-reperfusion phenomenon (affecting all organs) and a nonspecific activation of the systemic inflammatory response. During ischemia (phase of "no flow"), reduced oxygen supply is offset by lower metabolic needs. However, if cell metabolism remains requested or if the ischemia period is prolonged, the decrease of ATP synthesis leads to a plasma membrane depolarization, opening of voltage-dependent calcium channels sarcolemne, and fall of mitochondrial membrane potential. Schematically these phenomena result in an increase of intracytoplasmic calcium concentration responsible for cellular damage. Thus, it is during the phase of "no flow" that the first cell and tissue damage will occur. Reperfusion (phase of "low flow"), contemporary to the restoration of blood flow (created by chest compression, or spontaneous), is responsible for the formation of oxygen radical species, including superoxide anion (• O2-), hydrogen peroxide (H₂O₂) and the radical hydroxyl (• OH). The latter, particularly cytotoxic, carries for most functional and structural lesions causing cell death 4 . Indeed, it inactivates cytochromes, alters membrane transport proteins, and induces phenomena of membrane lipid peroxidation. In vivo demonstration was provided by studying plasma of out-of-hospital CA survivors, which induced acute and major endothelial toxicity, due to an acute pro-oxidant state that occurs within the cells, and illustrated by a decrease in antioxidant activity 5 . Disseminated vascular endothelium damages explain that this phenomenon of ischemia-reperfusion evolves toward systemic inflammation: production of cytokines (IL-1, IL-6, IL-8, TNF-α), complement activation, synthesis of metabolites of arachidonic acid, expression of leukocyte adhesion molecules by endothelial cells are all stimuli for activation, and chemotaxis of polymorphonuclear neutrophils at the origin of the inflammatory response. Sequestration of activated neutrophils in the lungs and other organs is a major engine for development of multiple organ failure. It is plausible that infectious lesions of digestive origin are involved, in relation to bacterial translocation from altered gastrointestinal mucosa. It would explain the high levels of plasma endotoxin observed in the most critically ill patients. This activation of the systemic inflammatory response associates with changes in coagulation, generating secondary endothelial damage, in turn responsible for thrombosis and increased capillary permeability 6 . Experimental models of CA support the pathophysiological hypothesis of worsening of visceral lesions that occur during the reperfusion phase and extend over the first hours, explaining the efficiency of some delayed therapeutic interventions (as therapeutic hypothermia). In humans, there is an activation of the inflammatory reaction associated with clinical features that are very similar to those observed during severe sepsis 7 . Coagulation abnormalities have been identified, involving a significant activation of coagulation factors, whereas endogenous anticoagulants (antithrombin, protein S and C) are decreased 8 . This intravascular coagulation can be implicated in the genesis of microvascular abnormalities, which in turn lead to more visceral lesions. Of note these coagulation abnormalities are particularly common in patients who die quickly from a post-resuscitation shock.

The postcardiac arrest syndrome presents a set of relatively stereotyped events 9 . Its intensity varies but is roughly proportional to the duration of "no flow" and "low flow," as well as difficulty encountered during initial resuscitation 10 . The cardiocirculatory failure usually dominates the clinical picture, even if it often leads to multiorgan failure (MOF). The decrease of cerebral perfusion, caused by these hemodynamic disturbances, might further impair the neurological prognosis of these patients.

The postcardiac arrest shock, initially described by Negovsky in 1975, is a mixed shock, with cardiogenic and vasodilatory components 2 and is characterized by a severe but reversible systolic dysfunction. Left ventricular dysfunction usually begins early, within minutes after ROSC, and is completely reversible within 48 to 72 hours. It takes the form of myocardial systolic and diastolic dysfunctions, even in the absence of coronary cause. The experimental evidence of this early, intense, and transient myocardial dysfunction after CA could be confirmed in humans. Indeed, Laurent et al. described the hemodynamic profile of postcardiac arrest shock 11 . In this study including 148 consecutive patients after a successfully resuscitated CA, 73 patients (49%) developed an early shock (within a median time of 8 hours after initial resuscitation) and were explored invasively with pulmonary artery catheter. Upon admission to the ICU, these patients exhibited a significant decrease in angiographic left ventricular ejection fraction, regardless of the etiology of CA. This shock was characterized by a low cardiac output and normal or low left filling pressures. Interestingly, in agreement with experimental data, the cardiac index improved quickly in patients with shock, from 2.1 l/min/m² (median value at the 8^th hour) to 3.2 l/min/m² at the 24^th hour, and up to 3.7 l/min/m² at the 72^th hour. The persistence of a low cardiac index was associated with early death, usually in a setting of MOF. But the circulatory failure does not appear to be related only to myocardial dysfunction. Indeed, despite the fast improvement of contractile function in survivors, this study also showed that vasopressors often had to be maintained until the 72^th hour, in association with important fluid infusion to maintain adequate filling pressures. Thus, these data support the existence of an early and intense myocardial failure, usually regressive within 48 hours, secondarily associated with a severe vasodilatation, as a result of the generalized inflammatory syndrome, itself well documented after CA 6 . Obviously, this has to be distinguished from the focal hypocontractility caused by a coronary thrombosis, which evolves on its own. The circulatory failure might also be favored by the occurrence of relative adrenal insufficiency, as observed during septic shock: clinical data show that approximately half of patients have relative adrenal insufficiency similar to what can be observed in septic shock. Among these patients, shock-related deaths seem to be more frequent in the group of those with adrenal insufficiency 12 .

The anoxic-ischemic neurological damages remain the leading cause of deaths occurring in patients initially resuscitated from CA. This neurological impairment can usually be detected from the third day after initial resuscitation. However, the dogma that the brain damage would be only due to injuries occurring contemporarily to circulatory interruption has been widely challenged during the past decade. Clinical studies that confirmed the efficiency of therapeutic hypothermia despite its "late" initiation, as regard to the onset of CA, support the existence of a worsening process during the reperfusion period. Indeed, neurological damage is initiated during circulatory interruption but also is accentuated during the reperfusion phase. Experimental and clinical studies show that the cerebral blood flow is decreased during the postcardiac arrest syndrome, associated with a decrease in cerebral oxygen extraction. The reduction of cerebral blood flow, which seems to be related to a transient increase in cerebral vascular resistance, will be corrected within 72 hours. Among survivors, the normalization of cerebral blood flow is accompanied by an increase in cerebral oxygen extraction, reflecting the recovery of the physiological coupling of these two parameters 13 . However, in patients with lethal brain injuries, the difference in coupling between cerebral blood flow and extraction appears gradually and finally results in a "luxury" perfusion.

Without specific and prompt treatment, postcardiac arrest shock usually leads to a multiple organ failure and early death. Acute renal and respiratory dysfunctions occur in approximately 40-50% of patients resuscitated from CA. Hypoxemia (a consequence of pulmonary edema, but also lung contusion, atelectasia, or aspiration), cardiogenic shock, acute renal failure, and liver failure may of course worsen the prognosis and delay neurological recovery. The digestive tract insult often is underestimated, due to the lack of bedside tool to investigate its function; however, its alteration should not be considered, both as a victim of the circulatory failure and as a second source of organ failure through translocation and major source of endotoxemia.

Many mechanisms that occur after CA make cardiopulmonary resuscitation survivors prone to infectious complications. Loss of airways' protection, coma, pulmonary contusion, emergency airway and vascular access, mechanical ventilation, and ischemia-reperfusion increase the risk for infection. The diagnosis is extremely complex, due to many clinical, biological, and radiological confounding factors. Moreover, therapeutic hypothermia plays a dual role by promoting infection and by impairing both host defences and usual infectious criteria. Recently developed biomarkers, such as procalcitonin, perform poorly in the assessment of septic complications 14 . We recently demonstrated that two thirds of CA survivors presented infectious complications during their ICU stay and may be favored by hypothermia. Despite increase in mechanical ventilation duration and length of stay, mortality and neurological outcome were not impaired by infectious events 15 . If antibioprophylaxis is not recommended, infectious complications should be cautiously tracked, especially when therapeutic hypothermia is employed.

The therapeutic management of the postcardiac arrest syndrome has two main goals: the initial treatment of shock and organ failures, and the optimization of cerebral protection. These two aspects are obviously very intricate, initial treatment having the essential interest to allow bringing the majority of patients to neurological evaluation, under hemodynamic stability conditions. The medical algorithm that we apply in our unit is displayed in Figure 2.

The existence of further cerebral damage during the reperfusion phase encouraged intense research on the benefit of treatments designed to limit the neurological consequences of the postcardiac arrest syndrome.

Achieving and maintaining a perfect homeostasis, particularly in terms of metabolism, are the major goals of post cardiac arrest management.

Immediately after CA, no clinical signs or investigations accurately predict the patient's outcome. During the first hours and days after ROSC, half of these patients will subsequently suffer from a postresuscitation disease, which can include a severe shock leading by itself to death. As mentioned earlier, shock should be treated without limitation during the very first hours and days to reach the adequate window for neurological evaluation. Finally, brain death occurs in approximately 10% of these patients a few days after CA because of cerebral edema 34 . Those patients are dead and life support will be stopped after completion of the organ-donation process.

After ROSC, accurate prognostication can only occur after 72 hours have elapsed from all factors that may alter neurological evaluation. This last point is of particular importance if sedation and neuromuscular blockade were used in conjunction with therapeutic hypothermia, because this treatment strategy must extend the waiting period 35 . Thus, if hypothermia is employed, it also is reasonable to use short-acting drugs. All other usual confounding situations (medications, hypothermia, and concomitant organ failure) must be resolved before any prognostication process.

Accurate assessment of prognosis is necessary to identify patients who will really benefit from intensive care and to avoid extending unnecessary treatments to those who have no reasonable chance of recovery. Prognostication should always be based on a rigorous clinical evaluation, which may include detailed neurological examination. Because there is no specific clinical sign that can predict outcome in the first few hours after ROSC, reliable prognostic factors and assessment of the ICU admission could be very useful. At this very early stage, a few elements are however available, and their accuracy are sometimes questionable. These elements can be grouped roughly into four main categories: intrinsic patient characteristics (age and comorbidities), parameters reflecting the quality of prehospital care ("no flow" and "low flow" durations), the etiology of CA, and finally, some of the patient's clinical characteristics:

- Intrinsic characteristics, including age, have a modest prognostic weight on early survival 36 ;

- The speed and quality of prehospital care logically have a major impact on the prognosis of patients. If "no flow" and "low flow" durations are valuable prognostic factors, with substantial mortality over respectively 5 and 15 minutes 36 , they cannot carry a final decision element. In addition, the accuracy of the information collected often is questionable in this context, which limits the scope of their prognosis ability.

- Concerning the cause of CA, asphyxia, and to a lesser degree, all extracardiac causes, are predictors of poor prognosis. In contrast, sudden cardiac death, occurring on a body properly oxygenated, is usually associated with better prognosis. However, once again, decisions cannot be made according to this oversimplistic split.

The absence of pupillary light response is not sufficiently predictive when it is sought on admission, due to a high false-positive rate, which can reach 30% 37 . Several additional clinical findings accurately predict poor outcome: myoclonus status epilepticus within the first 24 hours, absence of pupillary responses within days 1 to 3 after CPR, absence of corneal reflexes within days 1 to 3 after CA, and absent or extensor motor responses after 3 days 37 38 . Thus, the prognosis is invariably poor in comatose patients with absent pupillary or corneal reflexes, or absent or extensor motor responses 3 days after CA. On the opposite, single seizures and sporadic focal myoclonus do not accurately predict poor outcome. In the same way, although permanent status epilepticus is associated with a mortality approaching 100%, exceptions have been reported.

In addition to physical examination, various methods have been evaluated for assessing the neurological prognosis of comatose CA victims. Electrophysiological tests in coma consist of EEG and evoked/event-related potential studies. Somatosensory evoked potentials (SSEPs) are the best diagnostic method for predicting outcome in comatose patients. Bilateral absence of the N20 component of the SSEPs with median nerve stimulation recorded on days 1 to 3 or later after CA accurately predicts a poor outcome 38 . Conversely, the presence of the N20 response is not helpful to predict outcome as many patients who fail to recover can have preserved N20 responses.

Routine EEG examinations are not supported by the results of the published reviews but should be used if ongoing seizure activity is suspected. Generalized suppression pattern, burst-suppression pattern with generalized epileptiform activity, or generalized periodic complexes on a flat background are strongly but not invariably associated with poor outcome 38 .

Anoxo-ischemic cerebral insult is associated with the blood release of various biomarkers and the peak plasmatic level is thought to be correlated with the amount of neuronal death. A large serum peak of neuron-specific enolase and/or S-100 protein is highly specific but only moderately sensitive for predicting a poor neurological outcome 38 . The search for a simple, reliable, and readily available biological test remains an exciting challenge 39 , but clinical decisions with potentially irreversible consequences should never rely on a single marker.

Regarding routine neuroimaging techniques, CT-scan imaging adds no information unless stroke, bleeding, or trauma is suspected. Preliminary reports suggest that magnetic resonance imaging could be used to determine the prognosis of patients with diffuse cerebral anoxia 40 but warrants further investigations.

In a small number of distressing cases, patients regain spontaneous circulation but remain in a persistent vegetative state. It can be stated that continued existence of this state is not in the patient's best interest even if the only other alternative remains death. When this approach is shared both by care providers and relatives, end of life care must be considered: this consists in withholding/withdrawing all invasive and supportive care. Considering that most of these neurologically impaired patients are mechanically ventilated, the decision to withdraw therapies will nearly always include a statement about the pursuit of this assistance. When all conditions are joined, it is suitable to plan an extubation and the definitive withdrawal of respiratory assistance. This should be performed in conjunction with all comfort care that can be provided in such a situation (pain treatment, sedation). Decision to withdraw life-support treatment and the modalities of this withdrawal must be the conclusion of a collegial process between team and proxies, and all must be made to avoid the family's feeling of responsibility. Such decisions must obviously meet all the legal steps that exist in the country.

Major progress has occurred during the past 10 years in the initial management of successfully resuscitated CA: the development of a "chain of survival," the availability of automated external defibrillators in paramedical teams and public places, and the innovations in the resuscitation process gives hope for a further improvement in early survival. This will be accompanied by a steady increase in patients exposed to postcardiac-arrest syndrome. A better understanding of this syndrome undoubtedly transforms the last link of the survival chain: from an attitude of observation, the hospital management now becomes active and might condition in part the patient's progress in the short and long term. The primary objective of such care is to obtain survival with no or little neurological sequels. By influencing the vital and functional prognosis of patients, cerebral protection is now an essential part of the management of postcardiac-arrest patients. Currently, it relies mainly on therapeutic hypothermia. In the future, new pharmacological and nonpharmacological treatments should contribute to further improve the prognosis of these patients.

The authors declare that they have no competing interests.

NM and AC drafted the manuscript. All authors read and approved the final manuscript.