Prognostic indicators of survival and survival prediction model following extracorporeal cardiopulmonary resuscitation in patients with sudden refractory cardiac arrest

In this study, the key indicators for the survival of patients with ECPR were as follows: younger age (≤56 years), CPR duration up to 55 min, first documented cardiac rhythm without asystole, and any ROSC event during CPR. Our prediction model for survival after ECPR in the ED used these prognostic factors.

Several other studies reported prediction models for survival in veno-arterial extracorporeal membrane oxygenation (VA-ECMO) [21‐24]. The survival after VA-ECMO (SAVE) model comprises 12 pre-ECMO variables and was developed using data from a large international cohort with a validation process. Parameters such as age, weight, cause of cardiogenic shock (diagnosis), chronic renal failure, acute pre-ECMO organ failure, peak inspiratory pressure, duration of mechanical ventilation, pulse pressure and diastolic pressure before ECMO, and serum bicarbonate value before ECMO were used in this model [21].

The SAVE scoring system can be useful for patients with available pre-VA-ECMO parameters in the intensive care unit during hospitalization; however, most pre-ECMO parameters are not available for patients with OHCA and IHCA at the ED, as in our study.

Peter et al. reported an association between Sequential Organ Failure Assessment (SOFA) and survival in VA-ECMO; however, the calculation of this score requires laboratory data, as well as the type and dose of inotropic agents [23]. Pre-ECMO SAPS II was also reported to be a predictor of mortality in VA-ECMO; however, vital signs, laboratory data (including serum bicarbonate), and history of chronic disease are necessary for the calculation of SAPS II [24]. SAPS II and SOFA scores are used to predict mortality in the setting of critical illness, such as in the intensive care unit, and are not specific models for patients with ECMO. Most parameters for these scores are not available or cannot be accessed for patients with OHCA or IHCA shortly after admission to the ED.

Studies on SAVE, SAPS II, and SOFA have included patients who need VA-ECMO support for cardiogenic shock, not patients who need VA-ECMO for cardiac arrest [22‐24]. Even though the population in these studies included patients with cardiac arrest, those patients received ECMO support after IHCA, not OHCA. In our study, the prediction model for VA-ECMO survival focused on patients who received only ECPR at the ED. In all, 73.9% of the study population consisted of patients with OHCA. SAPS II may be a useful predictor in this study; however, the AUC of our model was higher than the AUC of the SAPS II scoring system. Moreover, the cutoff value of SAPS II was 91 points for ECPR, which is higher than the 80-point cutoff value for VA-ECMO in the study by Lee et al. [24]. Our prediction model may be useful for selecting and treating patients who need ECMO support during CPR but do not have patient information and laboratory data available at the ED.

The optimal cutoff age was 56 years for patients with sudden cardiac arrest at the ED or OHCA; there were no survivors older than 75 years. Other studies of prediction models for survival after ECPR for in-hospital cardiac arrest have identified a cutoff age of 66 years [16, 25]. This discrepancy may be due to the different patient characteristics between ED cardiac arrests and IHCAs, such as younger age, incomplete medical information, and pre-hospital variables for ECPR at the ED.

Some studies have reported that an age of <75 years predicts survival in CCPR [26, 27]. Maupain et al. also reported that old age was a risk factor for poor neurologic outcomes after CCPR for OHCA [28]. As the upper age limit for ECPR is generally set at 75 years, clinical data including very elderly patients with ECPR are rare [2, 14]. Younger patients with refractory cardiac arrest are good candidates for ECPR; however, the upper age limit for ECPR has not been validated [15].

The optimal time for implementation of ECPR, according to CPR duration, has been reported as follows: <30–60 min for survival [29, 30] and <55.5–80 min for favorable good neurologic outcomes [15, 31, 32]. A CPR duration of less than 55 min was an indicator of high likelihood of survival in our study. A meta-analysis by Debaty et al. [33] also concluded that shorter CPR duration was a prognostic factor in ECPR, associated with survival.

Reynolds et al. reported that the rate of favorable neurologic outcomes in OHCA decreased after 16 min of CPR and adequate CPR duration for probability of favorable neurologic outcome can be dynamically prolonged according to CPR-related variables, such as shockable arrest rhythm, witnessed arrest, and bystander CPR [10, 34]. As ECPR provides sufficient perfusion to vital organs, the window for an effective resuscitation duration can be extended [30, 31]. CPR duration is an indicator for the implementation of ECPR as well as an index to explain refractoriness to CCPR [35]. Providing ECPR to patients with refractory arrest within an optimal CPR duration is critical to achieve favorable outcomes.

The initial arrest rhythm (VF/pulseless VT or PEA, but not asystole) was found to be a predictive indicator of survival for ECPR patients. Other studies have indicated that a shockable arrest rhythm and witnessed arrest are prognostic indicators of survival and favorable neurologic outcome following CCPR [27, 28, 34, 36]. A shockable rhythm can be interpreted as a brief no-flow time or as an arrest of presumed cardiac etiology, which has also been reported to be a major factor of good outcomes in ECPR [12, 14, 30, 33]. Furthermore, studies by Park et al. and Kim et al. found that PEA was associated with survival and favorable neurologic outcomes in ECPR [15, 37]. When analyzed using three-categorized arrest rhythm, the prognostic factors for survival were as follows: age ≤56 years (OR 7.66, 95% CI 2.00–29.25, p = 0.003), CPR duration ≤55 min (OR 13.97, 95% CI 3.11–62.74, p < 0.001), VF/VT (OR 10.24, 95% CI 1.52–90.98, p = 0.037), PEA (OR 5.56, 95% CI 0.55–56.23 p = 0.146), and any ROSC event before ECPR (OR 8.3, 2.06–35.03, p = 0.003). The β coefficient of PEA was 1.72, whereas the β coefficient of shockable rhythm was 2.33. PEA was not a strong prognostic factor such as VF/pulseless VT, but one of predictors. The likelihood of a reversible etiology of arrest needs to be considered in the patients with PEA. ECLS plays a role as a bridge from refractory arrest to the correction of reversible causes of the arrest by ensuring adequate delivery of oxygenated blood until effective cardiac output has been restored. ECPR can be considered in the patients with PEA. However, implementation of ECPR for the patients with asystole is not recommended.

We found no differences between survivors and non-survivors with regard to pre-hospital variables, such as witnessed arrest and bystander CPR, in our univariate analysis. Witnessed arrest has been shown to be a prognostic factor for favorable outcomes in CCPR [26, 34]. However, this has not been clearly demonstrated for ECPR [12, 33, 38]. In this study, as in previous studies, prerequisites for ECPR included witnessed cardiac arrest or brief no-flow time; therefore, no association between witnessed arrest and survival could be identified.

The prediction model identified a ROSC event before ECPR as a predictor of survival in ECPR. CCPR achieves 25% of cardiac output, and high-quality CPR is important to achieve the necessary perfusion to major organs by compression [1, 35]. Any ROSC event can decrease low-flow duration with minimal cardiac output and increase the duration of perfusion.

In our univariate analysis, there were more patients without preexisting comorbidities in the survival group than in the non-survival group. The application of ECPR was usually limited to patients with a low Carlson comorbidity index (0 or 1). No differences in types or burden of comorbidities were detected [15].

The association between comorbidities and survival is not clear in ECPR, although Sǿholm et al. [39] found that comorbidity (low Charlson comorbidity index ≤2) was not independently associated with outcomes in CCPR. Similarly, comorbidity was not a prognostic factor in our study.

There were no differences between survivors and non-survivors with regard to in-hospital variables, such as serum lactate levels, and arterial pH as markers of inadequate tissue oxygenation and perfusion on ED admission. Lower lactate levels and higher arterial pH values have been reported among ECPR survivors [29, 33]. Serum hemoglobin levels on admission were associated with survival in this study. Hemoglobin levels can decrease due to outflow to the ECMO circuit and inflow from crystalloid solution of ECMO priming after ECPR. As decreased hemoglobin levels are associated with decreased oxygen delivery, impaired tissue function improvements in ischemic events during cardiac arrest may depend on the hemoglobin level [40, 41]. A mean arterial pressure ≥60 mmHg after ECPR was a dependent factor through hemodynamic optimization.