Predictors of In-Hospital Mortality after Recovered Out-of-Hospital Cardiac Arrest in Patients with Proven Significant Coronary Artery Disease: A Retrospective Study

In-Hospital mortality was 43.3%. Age and gender differences didn’t influence mortality. rOHCA non-survivors had a higher incidence of diabetes (42.3% vs 11.8%, p=0.01) and cerebrovascular disease (11.5% vs 0%, p=0.042). Diabetes also associated with a higher likelihood of multivessel disease (53% vs 93%, p=0.02, OR 12). With respect to other classic cardiovascular risk factors, no differences between groups were found (Table 1).

Regarding to pre-hospital variables (Table 1), rOHCA non-survivors were more likely to present with a non-shockable initial arrest rhythm (43.3% vs 8.8%, OR 7.5, p= 0.005), a longer low-flow time (17±13 min vs 22±11min, OR 3.3, p=0.003) and to be administered higher doses of adrenaline during ALS (3.6±3.0mg vs 1.8±2.3mg, OR 1.31 p=0.017). The best cut-off of low-flow time to predict survival was 11.5 min [area under the curve (AUC) by ROC analysis: 0.75].

There was no statistically significant difference in outcome related to the presence of ST-elevation, the extent of CAD and revascularisation strategy. However, non-survivors had a higher incidence of multivessel disease, larger troponin elevations and were less likely to receive PCI (Table 2).

The presence of moderate to severe left ventricular dysfunction, at hospital admission or during follow-up, didn’t influence mortality (Table 2).

Lactate levels in blood-gas analysis were higher in non-survivors: 7.9±5.1 vs 5.4±4.7mmol/L, OR 1.9, p=0.045. Non-survivors also had a lower lactate clearance in serial testing during the first 6 hours: 3.9±3.7 vs 2.2±1.5mmol/L, OR 2.2, p=0.048. Lower pH level at admission (7.17 ± 0.18 vs 7.27 ± 0.19, OR 0.05, p=0.046) was linked to the outcome. Reevaluation performed ~24h after admission revealed that the persistence of values out of the normal reference for pH and lactate (pH<7.35, OR 4.9, p=0.01 and lactates >2mmol/L, OR 3.4, p=0.04) signalled mortality (Table 3). The best cut-off point for lactate and pH at admission defined by ROC curve was 4.95mmol/L (AUC 0.65) and 7.25 (AUC 0.70), respectively.

In-hospital evolution in non-survivors was associated with serious complications (Table 2). rOHCA non-survivors had a significantly higher incidence of anoxic brain injury (57.7% vs 23.5%, p=0.007) and persistent hemodynamic instability (96% vs 38%, OR 4,4, p=0.001). As expected, the median duration of hospital stay was longer in the rOHCA survivors’ group (15, 6-22 days vs 3, 2-6 days, p<0.001).

In multivariate analysis of variables present at hospital admission, low-flow time superior to 12 minutes (p=0.006, OR 10.9) and non-shockable rhytmn (p=0.02, OR 6.9) independently predicted death. In multivariate analysis of in-hospital evolution variables, persistent hemodynamic instability independently associated with death (p=0.009, OR 24) (Table 4).

Survival according to the number of significant high-risk variables present at hospital admission and during hospital evolution was calculated using the Kaplan–Meier method. At hospital admission, initial rhythm, low-flow time, pH and lactates were the most relevant predictors and therefore were included in the score. Patients who had 0/4 criteria had 100% chance of survival till hospital discharge, 1/4 had 77%, 2/4 had 50%, 3/4 had 25%, respectively. Patients with all 4 criteria had 0% survival (log-rank p<0.001) (Figure 1). During hospital evolution, a pH<7.35 at 24h, lactates>2mmol/L at 24h, anoxic brain injury and persistent hemodynamic instability proved significant. Patients who had 0/4 criteria had 100% chance of survival till hospital discharge, 1/4 had 94%, 2/4 had 47%, 3/4 had 25%, respectively. Patients with all 4 criteria had 0% survival (log-rank p<0.001)(Figure 2).

Fig. 1

Fig. 2

Shockable initial rhythm, witnessed cardiac arrest, by-stander CPR and lower low-flow time increase the likelihood of survival. Multiple studies have proven that no-flow (time between the arrest and the beginning of BLS manoeuvres) and low-flow time (time from the beginning of BLS to ROSC) together with the quality of the cardiac massage are the major determinants of the success in attaining ROSC [15, 16, 17]. Likewise, an initial shockable rhythm predicts better outcomes since an effective treatment (i.e. shock/defibrillation) is immediately available to paramedics. Interestingly, Wah et al showed that an initial non-shockable rhythm, subsequently converted to a shockable rhythm was associated with better post-arrest survival and neurological outcomes when compared to a “continuous” non-shockable rhythm [18].

On the other hand, a non-shockable rhythm can be the initial arrest rhythm, in which case it tends to reflect a more chronic severe condition prior to the arrest and/ or is the consequence of longer delays in response that increase the probability of an initial shockable rhythm turning into a non-shockable one. An arrest at home/ private setting is less likely to be promptly rescued by professionals or people who can perform BLS manoeuvres. Therefore, there is a higher probability of deterioration of the patient’s status. However, regardless of the circumstances and primary pathophysiologic mechanism, a non-shockable rhythm strongly correlates with worse outcomes [19].

Adrenaline is one of the few drugs incorporated in advanced life support algorithms [20]. The vasoconstriction effect is mediated by α-adrenergic receptors and promotes increased coronary blood flow, leading to increased probability of ROSC but has no impact in survival to hospital discharge [21, 22, 23, 24]. Higher doses of adrenaline pre-ROSC associate with lower survival rate during hospital stay as they correlate directly with longer low-flow times [22, 23]. Furthermore, adrenaline was linked to ischemic lesions provoked by the vasoconstrictor effects on microvascular cerebral circulation and correlated to higher risk of poor neurological outcome [21, 25]. In the recent PARAMEDIC-2 trial >8000 patients were randomized to either adrenaline or placebo during ALS efforts. Investigators found that the use of adrenaline resulted in a higher rate of 30-day survival. However, there was no difference between the groups (adrenaline vs placebo) concerning the rate of a neurological favourable outcome since survivors in the adrenaline group had more frequently severe neurological impairment [25]. This powerful trial confirms the results of previous observational studies and questions the role of adrenaline use during OHCA.

The majority of studies refer an association between in-hospital survival and the performance of CA and PCI if necessary [1, 15, 26]. However, contrary to our study where only patients with sigCAD were included, the majority of these studies enrol patients with both coronary and non-coronary causes for OHCA. A study in OHCA patients with STEMI found that cardiogenic shock and multivessel disease were associated with mortality [9]. Our findings also revealed a tendency for more severe CAD and larger infarcts sizes (reflected by higher troponin elevations) in non-survivors although it didn’t reach statistical significance, pointing to the need of larger studies to prove its impact on survival

LVEF is considered an echocardiographic measure of myocardial dysfunction but its prognostic significance in OHCA patients is controversial and robust data is missing. Some observational studies report an association between LV dysfunction and survival [27, 28] while others [3, 29, 30], like ours, found that it doesn’t impact outcomes. Interestingly, Jentzer et al, showed that diastolic dysfunction is associated with in-hospital and long-term mortality, considering it a more important variable in predicting patient outcomes [31].

Blood-gas analysis provides critical information after rOHCA and unfavourable outcomes correlate with higher lactate and lower pH levels at admission and during hospital stay, reflecting a longer body exposure to ischemia [32, 33, 34]. However, the cut-off points of blood lactate and pH levels are not well established. pH is influenced by CO2 clearance, HCO3- and anion gap which includes lactates, hence its variation may reflect more broadly on the status of vital organs. pH levels can therefore be a more specific predictor of outcome when compared to lactate levels, as seen in Momiyama et al [34]. In our study, we went further in exploring the importance of metabolic alterations. Interestingly, we found that the inability to achieve metabolic balance, expressed by the maintenance of altered values at the 24h blood-gas analysis, strongly correlated with mortality. This persistent metabolic derangements reflect circulatory failure with subsequent impaired regulation of homeostasis, both of which classically associated with dismal prognosis.

In-hospital complications are frequent in OHCA patients. Anoxic brain injury is the most frequent cause of death after OHCA occurring in up to 66% of patients [35]. Current guidelines recommend using information from clinical, biochemical and imaging tests performed >72h after the arrest to more accurately define irreversible neurological damage[36]. Using this criteria, anoxic brain injury was present in 38% of our patients and powerfully signalled mortality. Accurately defining a poor neurologic outcome in the OHCA population is of crucial importance to avoid untimely withdrawal of care.

Hemodynamic instability was present in 63% of the patients included in this study and independently related to mortality. Clinically, it can manifest as persistent hypotension, low cardiac index and dysrhythmias and is a common complication of OHCA. The mechanisms leading to persistent instability include hypovolemia, microcirculatory deregulation and myocardial dysfunction [37]. Hypovolemia is common and can be caused by hypothermia-induced diuretic effect, by active blood loss or by loss of plasmatic fluid to extravascular compartments. The typical reperfusion injury after OHCA is caused by marked inflammatory response, leading to microcirculatory dysfunction and impaired regulation of vasoconstriction [38]. These mechanism not only aggravate ongoing ischemia and multiorgan dysfunction but also potentiate each other, leading to a vicious circle that might culminate in death.

The management of post-arrest patients is challenging. Although there are multiple risk-stratifying scores in this population, its use is not widespread.Limiting factors include their complexity and the lack of applicability in particular scenarios. The acute physiology and chronic health evaluation (APACHE) II score was developed in 1985 and is still one of the most widely used score in intensive care units (ICU) [39]. Although more recent versions were published, they are not freely available (under copyright control). The APACHE II score uses a range of simple demographic, clinical and biochemical variables to determine prognosis. This score was validated for the general ICU population and there are some challenges when it comes to its direct application to OHCA patients: First, it doesn’t include OHCA as a variable, which is of critical importance to the patient outcome. The score applicability in post cardiac-arrest was tested in a prospective study that concluded it was a poor predictor of the outcome [40]. Second, risk estimation calculations were done in 1985 and are probably less accurate in estimating prognosis in patients treated nowadays.

The OHCA score propectively validated the use of 5 variables (shockable rhytmn, no-low time, low-flow time, creatinine and lactates) to predict mortality after OHCA [41]. In this study, logarithmic transformation of continuous variables was performed to avoid the “class jump phenomenon”. The authors refer that this method contributed to the good performance of the score (AUC 0.88). However, the score is not readily available in standard online medical calculators and the need to use mathematical functions for each variable makes it less user-friendly.

Skrifvars et al. compared the performances of the APACHE III and the OHCA scores to predict outcome following OHCA. The authors concluded that both scores offer moderate predictive accuracy (AUC for OHCA score: 0.77; for APACHE III score: 0.71) but correlate weakly with each other [42].

In the present study, we first corroborated the usefulness of classical variables in this rOHCA subgroup;-Then, we tried to refine its clinical applicability. The use of the 4 hospital admission variables, namely pH, lactates, low-flow time and initial rhythm,lead to a risk of death stratification of patients at admission. The presence of 4/4 at hospital admission estimated a 100% mortality in our series. Likewise, during in-hospital evolution, having persistent hemodynamic instability, anoxic brain injury and the inability to achieve metabolic balance clearly signalled a high mortality risk. Our simple approach might contribute to the identification of patients at high risk for adverse outcomes. However, larger studies are needed to validate this findings.

Taken together, these data reinforces the notion that a “one size fits all” score will hardly be found for OHCA. Therefore, prognostic assessment in these patients needs to be individualized.

We faced some challenges along our work. The retrospective nature of this study and the fact that it was based on medical records resulted in missing data to some extent, making it more prone to be influenced by bias and confounding factors. Our inclusion criteria were defined to better portray a group of patients with specific features. However, they inevitably led to a smaller sample size, that limited the study power to detect variables that may be associated with in-hospital mortality.