Intensive care unit mortality after cardiac arrest: the relative contribution of shock and brain injury in a large cohort
- First Online:
- Cite this article as:
- Lemiale, V., Dumas, F., Mongardon, N. et al. Intensive Care Med (2013) 39: 1972. doi:10.1007/s00134-013-3043-4
Brain injury is well established as a cause of early mortality after out-of-hospital cardiac arrest (OHCA), but postresuscitation shock also contributes to these deaths. This study aims to describe the respective incidence, risk factors, and relation to mortality of post-cardiac arrest (CA) shock and brain injury.
Retrospective analysis of an observational cohort.
24-bed medical intensive care unit (ICU) in a French university hospital.
All consecutive patients admitted following OHCA were considered for analysis. Post-CA shock was defined as a need for infusion of vasoactive drugs after resuscitation. Death related to brain injury included brain death and care withdrawal for poor neurological evolution.
Measurements and main results
Between 2000 and 2009, 1,152 patients were admitted after OHCA. Post-CA shock occurred in 789 (68 %) patients. Independent factors associated with its onset were high blood lactate and creatinine levels at ICU admission. During the ICU stay, 269 (34.8 %) patients died from post-CA shock and 499 (65.2 %) from neurological injury. Age, raised blood lactate and creatinine values, and time from collapse to restoration of spontaneous circulation increased the risk of ICU mortality from both shock and brain injury, whereas a shockable rhythm was associated with reduced risk of death from these causes. Finally, bystander cardiopulmonary resuscitation (CPR) decreased the risk of death from neurological injury.
Brain injury accounts for the majority of deaths, but post-CA shock affects more than two-thirds of OHCA patients. Mortality from post-CA shock and brain injury share similar risk factors, which are related to the quality of the rescue process.
KeywordsOut-of-hospital cardiac arrestPost-cardiac arrest syndromeMortalityShockCare withdrawal
Out-of-hospital cardiac arrest (OHCA) is a leading cause of death in Western countries, with around 100,000 people treated for OHCA in the USA each year. The prognosis for these patients remains poor: based on a recent international meta-analysis, 7.6 % of treated OHCA patients survive to hospital discharge . Among successfully resuscitated patients admitted to intensive care unit (ICU) after OHCA, less than 30 % will be discharged from hospital [1, 2]. This early mortality after resuscitation is caused by postresuscitation circulatory failure (mainly due to systemic ischemia–reperfusion) and postanoxic brain injury [3–8].
While underlying causes of OHCA and ischemia–reperfusion have been studied widely [8, 9], little is known about post-cardiac arrest (CA) shock. Although its pathophysiology has been well described , including cardiac dysfunction [11, 12] and vasoplegic shock [13, 14], risk factors associated with post-CA shock and its contribution to mortality have never been specifically assessed. In most countries, the presence of severe postanoxic neurological injury commonly results in a decision to withdraw life-sustaining treatments, and this mechanism may account for half of deaths after CA [6, 7]. A few studies have specifically focused on mode of death after CA [3, 15]. However, these were performed on small cohorts, and the relation between post-CA shock and death was not assessed. Lastly, one study  was conducted before the wide application of therapeutic hypothermia (TH), which has been shown to be an effective neuroprotective therapy [16, 17].
Early identification of patients at high risk for post-CA shock could help to direct specific treatment . The first objective of this study is thus to assess the incidence and risk factors of post-CA shock and its consequences. The second aim is to identify the potential predictive factors for death related to post-CA shock or brain injury in a large cohort of OHCA patients.
Patients and methods
We performed a retrospective study of our database, between January 2000 and December 2009, in our closed 24-bed medical ICU of the Cochin Hospital (Paris, France). All survivors and nonsurvivors’ next of kin were informed that data were entered into a database in order to be used for clinical research. According to French law, our institutional review board waived the need for written informed consent.
All information from patients admitted after OHCA was routinely collected prospectively in a database, according to Utstein style .
All patients included in the database were screened for participation. Patients who died within 1 h after admission were not considered in the analysis.
OHCA was considered of cardiac origin (acute coronary syndrome and/or cardiac arrhythmia), respiratory origin (i.e., pulmonary embolism or acute respiratory failure), neurological origin, or other miscellaneous (sepsis, anaphylaxis, metabolic disorders). Unknown cause was documented when exhaustive investigations [coronary angiogram, brain and thorax computed tomography (CT) scan, electrocardiogram, transthoracic echocardiography, routine biology or cerebrospinal fluid analysis, toxicological screening, necropsy] remained inconclusive.
Post-CA shock was defined as the need for continuous norepinephrine or epinephrine infusion to maintain mean arterial pressure above 60 mmHg for more than 6 h following restoration of spontaneous circulation (ROSC), despite adequate fluid loading.
Neurological outcome was assessed daily by ICU physicians until death or ICU discharge according to a strictly controlled procedure described in the ESM .
Finally, the cause of death was defined for each nonsurvivor as related to post-CA shock, when death occurred as a direct consequence of shock (including subsequent multiorgan failure), or related to neurological injury if this led to withdrawal of life-sustaining treatment (WLST) or brain death. Only deaths occurring during the ICU stay were considered in this study. All patients who died after a WLST decision concomitant with abnormal neurological clinical examination were considered as patients who died from neurological injury.
We summarize categorical variables as proportions and nonparametric continuous variables as median and interquartile range, and we compared them using Pearson’s chi-square test or Fisher’s test, if appropriate. We describe continuous variables by means and their standard deviation (SD), and used Student t test for comparison.
Firstly, we assessed the factors predicting postresuscitation shock using univariate analysis and then a multivariate model using logistic regression using all variables. Secondly, we analyzed the mode of death (i.e., post-CA shock or neurological injury) independently of post-CA shock onset. We included all potential confounders in a multinomial logistic regression to identify factors associated with one or the other mode of death.
Instead of excluding cases with missing covariates, we conducted a sensitivity analysis first using multiple imputations to incorporate all subjects in the fully adjusted model and second excluding those with covariates missing from the multivariable model.
Finally, to assess the potential influence of time, the study period was divided into three time intervals: before 2003, corresponding to the “prehypothermia era”; between 2004 and 2006, the period following the publication and dissemination of preliminary international guidelines; and after 2007, when hypothermia was used routinely for treatment of OHCA [18, 22].
Odds ratios (OR) were calculated with their 95 % confidence interval (CI). All tests were two-sided, with p ≤ 0.05 considered statistically significant. We performed analyses using STATA/SE 11.2 software (College Station, TX, USA).
The median age was 59 (SD 16) years, and 842 (73 %) patients were male. The cause of OHCA was cardiac in 674 (59 %) patients, respiratory in 242 (21 %) patients, neurological in 91 (8 %) patients, and of miscellaneous or unknown etiology in 145 (13 %) patients. CA occurred at home for 771 (67 %) patients, and CPR was provided by bystanders in 464 (41 %) cases. First cardiac rhythm was shockable for 654 (57 %) patients, and the median time interval between collapse and ROSC was 20 (12–31) min. TH was performed in 764 (66 %) patients.
Post-CA shock: incidence and risk factors
Characteristics of patients with and without post-cardiac arrest shock
Post-cardiac arrest shock, n = 789
No shock during ICU stay, n = 363
Male gender, n (%)
Age, years (SD)
Location of OHCA, n (%)
Bystander CPR, n (%)
Shockable rhythm, n (%) (VF/VT)
Time from collapse to ROSC (min)
Etiology of cardiac arrest
Other or unknown
Admission lactate (mmol/L)
Admission creatinine (µmol/L)
Therapeutic hypothermia, n (%)
ICU mortality, n (%)
Multivariate analysis of predictive factors associated with occurrence of post-cardiac arrest shock during ICU stay
OR (95 % CI)
Time from collapse to ROSC (per category increase)
Arterial lactate value (per category increase)
Creatinine value (per category increase)
Mode of death
During ICU stay, 768 (66 %) patients died (Fig. 1). A comparison between survivors and nonsurvivors is summarized in Table 1 of the ESM. Survivors were younger, with shorter resuscitation duration and higher bystander CPR rate. Survivors were also more likely to have had a shockable rhythm, a cardiac cause, and lower lactate and creatinine values.
On multivariate analysis (data not shown in tables), independent factors associated with ICU death were longer time from collapse to ROSC (p < 0.001), age (p < 0.001), and higher lactate value on admission (p < 0.001). Although there was a trend in our study, post-CA shock onset was not associated with higher risk of death (p = 0.072).
Most patients died from WLST after 6 days (Fig. 2), but for 40 patients, death occurred within the first 4 days after ICU admission. Among this last subgroup, brain death was diagnosed in eight patients, and the other patients were considered to have too poor neurological prognosis.
Only 5 % (n = 51) of patients died beyond 14 days. In this subgroup, there were 40 (78 %) males, with a median age of 59 (49–71) years. No flow was 4 (0–10) and low flow was 15 (8–25) min. Among them, 32 (62.7 %) had post-CA shock during ICU stay. Death was related to WLST for 45 (88 %) patients and septic shock for 6 (12 %) patients with neurological injury.
Multivariate analysis according to mode of death
Deaths related to post-cardiac arrest shock, n = 269
Deaths related to neurological injury, n = 499
OR (95 % CI)*
OR (95 % CI)*
Cardiac arrest location (public area versus home)
Time from collapse to ROSC (per category increase)
Admission lactate (per category increase)
Admission creatinine (per category increase)
After multiple imputations, the model and associations remained very stable (Table 2, ESM).
Impact of time on mode of death
We investigated the influence of therapeutic hypothermia on the mode of death timeline. There was no change in the proportion of deaths related to neurological injury and post-CA shock during the three predefined periods (Fig. 2, ESM). The proportion of patients with post-CA shock was similar in each period (respectively, 71, 60, and 74 %).
In our cohort, 64 % of patients died after ICU admission, a rate that is consistent with studies from other Western countries [1, 6, 7, 23]. Laver and colleagues authored a study dedicated to ICU cause of death after CA. They reported that shock and multiorgan failure were responsible for 31.3 % of deaths in 113 OHCA patients, a proportion that is consistent with our findings . A more recent Swedish study (162 OHCA, 47 % survival rate) reported three modes of death: brain injury (71 % of nonsurvivors), shock (16 %), and miscellaneous etiology (13 %) . Nevertheless, the small size of the cohorts, the lack of details about shock, the absence of TH in one study , and the lower survival rate usually reported after CA called for a new evaluation of a large unselected cohort. The present study reports the largest analysis performed in a post-cardiac arrest registry.
To further understand the post-CA syndrome, we investigated respective causes of death. In this retrospective analysis of a large observational cohort, we found that post-CA shock occurred in around two-thirds of OHCA patients. Even if brain injury was responsible for the majority of deaths, post-CA shock was responsible for one-third of ICU deaths. The relative contributions to mortality made by post-CA shock and brain injury were the same in each of the three study periods. Finally, we found that mortality from post-CA shock and brain injury shared similar risk factors.
Post-CA shock was described 40 years ago , but the hemodynamic features were reported in the landmark study of Laurent et al. . Despite the lack of a consensus definition, it is well accepted that this shock has cardiogenic and vasodilatory components. It is characterized by severe, early, but reversible diastolic and systolic cardiac dysfunction, which occurs even if the cardiac arrest had a noncardiac cause. Severe vasodilation is a consequence of the generalized inflammatory syndrome, which is well documented after CA . In previous studies, the proportion of patients requiring vasoactive drugs after ICU admission because of hemodynamic instability ranged from 27 %  to 62 % . In our large series of patients, use of vasoactive drugs was more common than in previously published studies. This might reflect the specific characteristics of our population: the commonest primary cause of CA was cardiac, which is more commonly associated with shock . The severity and potential reversibility of cardiac dysfunction highlight the importance of understanding the risk factors associated with post-CA shock [11, 12]. We determined that the factors independently associated with post-CA shock were mainly prehospital (duration of resuscitation), as well as arterial lactate values at admission. Lactate value at admission roughly reflects the quality of resuscitation during OHCA. Likewise, we found that this marker was associated with post-CA shock during ICU stay. Finally, an elevated lactate level could be an alarm for “cryptic shock” detection, and could result in earlier detection of circulatory compromise. Adrie and colleagues  previously showed that longer resuscitation duration was associated with lower survival rate, but our study is the first one to show a relation between long resuscitation duration and occurrence of postresuscitation shock. Shock prediction at time of ICU admission may enable establishment of very close hemodynamic monitoring, or tailoring of specific therapeutic interventions such as high-volume hemofiltration , steroid administration  or cardiac assistance . In our study, therapeutic hypothermia was not associated with onset of shock. This should be considered in light of the two landmark studies [16, 17], which did not include patients with shock. Increasing data on the safety of mild hypothermia in the presence of postresuscitation shock support our practice [28–30].
Post-CA shock was an early cause of death: almost 50 % of patients exhibiting shock died from this cause. However, we confirmed that post-CA shock was not associated with a higher risk of death on multivariate analysis, even for patients treated with TH [29–33, 35]. In our study, factors independently associated with death related to postresuscitation shock were prehospital resuscitation duration, blood lactate value, and age. This emphasizes the importance of optimal prehospital management . On the other hand, shockable rhythm and TH were associated with a lower risk of death related to post-CA shock. The protective role of a shockable rhythm probably reflects its well-established better prognosis, but our findings in relation to TH are new and of paramount importance . Our study also highlights the role of TH in a large population of unselected patients, particularly when post-CA shock occurs. This is consistent with the protective effect of hypothermia in patients with post-CA shock, which has been described in previous studies [30, 34, 35]. Nevertheless, we could not show a definite relationship between TH and decreased risk of death related to shock because, in more hemodynamically unstable patients, TH would have been stopped. Identification of high-risk patients soon after admission might enable implementation of new therapies to treat reversible cardiogenic shock, such as extracorporeal life support [36, 37]; this may enable more patients to survive this early phase and hopefully achieve favorable neurological recovery.
Factors associated with death from neurological injury were mainly prehospital. This finding is consistent with data from France and other countries [1, 7]. In countries where prehospital care could be delivered earlier, overall mortality was lower in the studies [38, 39]. Thus, our study reinforces the need for improving prehospital care.
Progressive implementation of post-CA therapeutic hypothermia during the study period was not associated with a decrease in neurological cause of death in our cohort. Contrary to the two pivotal studies, which included only ventricular fibrillation cardiac arrests [16, 17], all-rhythm cardiac arrests were included in this study, and a recent study showed no benefit from TH following CA with nonshockable rhythms . Nevertheless, changes in case mix in our study may be one of the reasons why we did not show an association between TH and reduced death from neurological injury over time. In our study, death from neurological injury occurred mainly after the first 3 days. Recognition of the poor reliability associated with early neurological assessment  means that WLST decisions are now being made later than in previous study . Interestingly, we did not observe a change in the distribution of mode of death over the three time periods, even though this covered phases before and after TH implementation.
Finally, we documented that 12 % of patients met the criteria for brain death. This result is consistent with that of Adrie et al. . These findings are of utmost importance, given the severe shortage of organs for transplantation. These findings further support aggressive resuscitation of OHCA patients, because even if their neurological outcome is poor, there is potential for organ donation. Regrettably, as was the case in Adrie’s study , there were too few patients to enable analysis of the characteristics associated with this outcome.
Some limitations of our study merit discussion. The single-center design means that changes in local practice could have influenced our findings significantly. Tools for prognostication were modified in 2005, in particular the introduction of somatosensory evoked potentials . Nevertheless, end-of-life decisions were strictly controlled by a legal framework  (Fig. 1, ESM). Moreover, our ICU is a referral CA center, caring for more than 100 patients each year. Second, the study was performed over a long period. The single-center design meant that our diagnostic and therapeutic strategy was consistent and followed international guidelines . Third, our definition of post-CA shock could be debated. We chose to embrace a pragmatic and real-life definition of vasopressor requirement instead of a more subtle definition . Moreover, we did not perform autopsy, which could have clarified some diagnoses . Fourth, in our Utstein-style database we did not record time delays between end-of-life decisions, extubation, WLST, and death. This would have provided better insight into WLST in this setting. Moreover, this information is of paramount importance in countries where use of organs from donation after cardiac death is permissible . Fifth, our findings apply only to OHCA, the distribution of mode of death being quite different following in-hospital CA . Also, we could not exclude some potential residual confounding factors to explain risk factors associated with death from post-CA shock or neurological injury; For example, many patients who died from post-CA shock had not been extensively assessed for neurological injury, and we cannot exclude that some of them would have secondarily died from neurological injury. Finally, the observational design implies data missed from collection. However, even after taking into account different methodological approaches including multiple imputations, our findings remained very stable.
In conclusion, neurological injury is responsible for the majority of deaths after successfully resuscitation after OHCA. However, this study emphasizes the high incidence of post-CA shock and its impact on ICU mortality. Further studies should focus on optimizing treatment of post-CA shock.
Conflicts of interest