The Utility of Risk Scores in Postcardiotomy Extracorporeal Membrane Oxygenation

Introduction

Postcardiotomy shock (PCS) is the cardiogenic shock that occurs immediately following cardiac operations despite high-dose inotropes and intraaortic balloon pump support and is seen after 3–5% of cardiac operations. ¹ The use of a venoarterial extracorporeal membrane oxygenation (ECMO) in the postcardiotomy shock setting (PC-ECMO) can be mandatory and life-saving. ² It allows myocardial unloading in the period of ischemia-reperfusion injury and stunning while preserving end-organ perfusion. The need for PC-ECMO arises in 0.5–1.5% of cardiac surgery patients.^1,2

Despite advances in ECMO care, not all patients with PC-ECMO can be weaned and the survival to discharge in these patients is reported at a rate of 25–44%.^3,4 Risk stratification for patients under PC-ECMO is currently challenging. ⁵ Decision-making for initiating ECMO for PC patients is difficult and an effective and validated scoring tool for predicting survival is a necessity.

The EuroSCORE II and the Society of Thoracic Surgeons (STS) score are established scoring systems used widely for preoperative risk calculation in cardiac surgery.^6,7 The variables in these scores are limited to preoperative factors and lack survival prediction for ECMO. ⁸ Other risk scores have been developed and proposed for ECMO patients of various indications. The Survival after veno-arterial-ECMO (SAVE) Score was developed by the ELSO at The Alfred Hospital, Melbourne to assist in mortality prediction in adult cardiogenic shock patients. ⁹ A modification to the SAVE score was suggested using lactate as an additional parameter. ¹⁰ The vasoactive inotropic score (VIS) is a quantified sum of the inotrope doses received by a patient which was shown to be prognostic in cardiac surgery and ECMO patients. The aim of this study was to assess and compare the discriminatory ability of the different available risk scores for mortality and successful weaning of PC-ECMO patients.

Methods

Patients aged >18 years undergoing coronary artery bypass and/or valve surgery and implanted an ECMO intraoperatively for postcardiotomy shock between January 2017 and June 2022, were retrospectively included. Patients of aortic surgery or other cardiac procedures and patients who were implanted an ECMO in the intensive care unit were excluded. The study was performed in an ELSO registered tertiary center for cardiac surgery. Approval was obtained from the associated ethics committee (HNEAH-KAEK 2021/279-3342) for this study. The outcomes of interest were 30-day mortality, successful weaning from ECMO, and survival to discharge. In the patients, ECMO was implanted in case of failure from weaning off cardiopulmonary bypass intraoperatively despite metabolic and inotropic management with three inotropic agents and IABP no later than the first three hours of attempted weaning. The follow up of ECMO patients were performed according to ELSO Guidelines. ¹¹

Results

During the study period, 7342 patients underwent coronary bypass and/or valve surgery in our institution. Of these, 115 patients were implanted an ECMO for postcardiotomy shock at the end of the operation. Nine patients died in the operative room were excluded from analysis as all risk scores could not be calculated for these patients. The characteristics of the study patients with PC-ECMO are summarized in Table 1.

OPEN IN VIEWER

Table 1.

Characteristics of PC-ECMO patients.

	n = 106
Patient characteristics
Age	56.3 ± 12.7
Female gender	34 (32.1%)
Hypertension	46 (43.4%)
Diabetes mellitus	37 (34.9%)
Body mass index	27.7 ± 5.7
Chronic renal failure	5 (4.7%)
Chronic obstructive pulmonary disease	13 (12.3%)
Peripheral arterial disease	2 (1.9%)
Preoperative left ventricular ejection fraction	46.7 ± 13.1
ECMO characteristics
Central cannulation	41 (38.7%)
Central cannulation with LV unloading	39 (36.8%)
Peripheral cannulation	26 (24.5%)
IABP use with ECMO	12 (%11.3)
Revision for bleeding/Tamponade or peripheral ischemia in the first 24 h	7 (%6.6)
Risk scores
Euroscore II	2.83 ± 2.29
STS score	1.90 ± 1.69
APACHE II	41.40 ± 21.52
SAVE	−5.30 ± 3.55
Modified SAVE	−5.96 ± 3.67
VIS	39.79 ± 14.28
ECMO outcomes
Weaning	55 (51.9%)
Survival to discharge	28 (26.4%)

The risk scores were calculated for each patient. Each risk score was separately evaluated for its association with 30-day mortality. (Table 2) The Euroscore II and STS scores were not associated significantly with 30-day mortality, whereas The SAVE score, the modified SAVE score, APACHE II, and VIS all had significant associations with 30-day mortality on PC-ECMO. A 10-point increase in APACHE II and VIS scores increased the risk of 30-days mortality by 1.19 and 1.28, respectively.

OPEN IN VIEWER

Table 2.

Cox proportional hazard model for 30-day mortality.

Risk score	HR (%95 CI)	p
Euroscore II	0.950 (0.862–1.047)	.300
STS	1.094 (0.960–1.248)	.177
SAVE	0.958 (0.934–0.983)	.001
Modified SAVE	0.954 (0.929–0.979)	<.001
APACHE II	1.017 (1.007–1.028)	.001
VIS	1.025 (1.010–1.041)	.002

The scores of interest were compared using ROC curves constructed for ECMO weaning (Figure 1) and survival to discharge (Figure 2). The comparison of the AUCs for ECMO weaning are presented in Table 3 and for survival to discharge in Table 4. In survival to discharge, VIS had the highest AUC with 0.844 followed by the APACHE II score with 0.727. In weaning from ECMO, APACHE II had the best AUC with 0.715 followed by VIS with 0.698. The SAVE and modified SAVE scores performed similarly for survival. The Euroscore II and STS had the poorest discriminatory ability for both weaning and survival to discharge.

OPEN IN VIEWER

Figure 1.

ROC curves of different risk scores for ECMO weaning.

OPEN IN VIEWER

Figure 2.

ROC curves of different risk scores for survival to discharge.

OPEN IN VIEWER

Table 3.

Performance of risk scores for ECMO weaning.

Risk score	AUC	p	Sensitivity	Specificity	Youden index	PPV	NPV
VIS	0.698	.002	0.935	0.419	28.25	0.632	0.857
Apache II	0.715	<.001	0.843	0.618	37.03	0.672	0.810
Modified SAVE	0.611	<.001	0.686	0.698	−30	0.686	0.698
SAVE	0.637	.002	0.745	0.600	−36	0.633	0.717
Euroscore II	0.569	.312	0.667	0.491	1.815	0.548	0.614
STS	0.510	.388	0.549	0.673	1.775	0.609	0.617

OPEN IN VIEWER

Table 4.

Performance of risk scores for survival to discharge.

Risk score	AUC	p	Sensitivity	Specificity	Youden index	PPV	NPV
VIS	0.698	<.001	0.889	0.765	28.25	0.941	0.619
Apache II	0.715	<.001	0.731	0.750	37.03	0.891	0.500
Modified SAVE	0.611	<.001	0.684	0.643	−37	0.867	0.455
SAVE	0.637	.002	0.667	0.714	−36	0.867	0.435
Euroscore II	0.569	.909	0.628	0.500	1.13	0.778	0.326
STS	0.510	.428	0.474	0.679	1.78	0.804	0.317

Discussion

Postcardiotomy shock patients represent a specific patient group who are difficult to assess for risks and benefits. Where the traditional cardiac surgery risk scores are lacking, newer scores such as the VIS and the SAVE score can perform better in predicting successful weaning and discharge. Based on our results, VIS, Apache II, modified SAVE, and the SAVE scores are associated with ECMO outcomes in postcardiotomy shock.

There is currently no risk score designed for predicting survival for PC-ECMO patients. Of the risk scores developed using the ELSO registry, the only score for venoarterial ECMO in the adult population is the SAVE score. ⁹ It includes 11 variables pertaining to indication, age, weight, cardiac, respiratory, renal, hepatic, and neurological parameters before ECMO implantation and was developed from a patient population that included cardiogenic shock of any cause. In separate single center studies, the modified SAVE score which adds points for low lactate levels was shown to have better discrimination for mortality in patients with cardiogenic shock in the emergency setting.^9,10,16 The original and the modified SAVE scores performed better than the classical cardiac surgery risk scores for both mortality and weaning in our cohort. The lactate modification of the SAVE score was slightly superior in mortality prediction. The lactate dynamics during the early stages of ECMO treatment are shown to be influential in survival, which may explain the increased discriminatory power with the incorporation of lactate levels into the SAVE score. ¹⁷

ECMO patients are heterogeneous and while a risk score may perform better for acute cardiogenic shock patients they may perform poorly in other patient populations. Among cardiac patients supported with an ECMO, a postcardiotomy indication almost doubles the risk of mortality. ¹⁸ With the evident need for an effective algorithm, newer risk scores and scores originally used in general ICU patients have been investigated for their value in ECMO patients. It is difficult to assess the state of the patients at the moment of ECMO implantation in postcardiotomy patients as preoperative values do not completely reflect the state at the end of cardiopulmonary bypass. Most end-organ functions including the renal function or the neurological state are unknown at the moment of ECMO initiation. Clinicians rely on hemodynamic measurements and operative laboratory results as well as the patients’ preoperative status to estimate the risks and benefit on an ECMO run. Therefore, it is expected that risk scores that solely rely on preoperative parameters do not function well for PC-ECMO patients. While a need for guidance is highest before ECMO initiation, scores that rely on parameters after ECMO initiation perform better as the early state after ECMO initiation maybe a better indicator of success than the patient’s preoperative state.

The patient’s hemodynamic measurements, inotrope requirements, lactate levels and biochemical results are incorporated by the SAVE, modified SAVE, VIS, and APACHE II scores which performed better in the PC-ECMO patient population than the STS or EuroSCORE II. In a study that compared risk scores in PC-ECMO patients, the best discrimination was achieved by the SAVE score for short and long term mortality, with a reported AUC of 0.62 and 0.67 respectively. ¹⁹ The results from a multicenter study performed in large volume European studies also found SAVE, modified SAVE, and APACHE II scores to significantly predict mortality, while APACHE II had the best discrimination with an AUC of 0.716. ²⁰ Our findings were similar in that the SAVE and the modified SAVE scores showed moderate discrimination ability with AUCs of 0.672 and 0.695 which limits their clinical utility for PC-ECMO patients.

The APACHE and VIS scores had a more satisfactory discrimination for survival to discharge with AUCs of 0.727 and 0.844, respectively. In a single-center study, Pre-ECMO VIS >60 has been associated with approximately 2.8 times increased risk of 30-day and 3-year mortality. ²¹ In our cohort, a 1.28 times increase in risk with a 10 point increase in 24th hour VIS could be demonstrated. Current data suggest that ECMO treatment is more successful with early initiation of ECMO support before the onset of organ injury or uncontrollable bleeding. ²² Therefore, the pre-ECMO VIS may not be applicable when the patient is started on ECMO intraoperatively. The total dose of inotropes after 24 h of ECMO treatment parallels the cardiac and systemic functions of the patients under ECMO treatment and provided satisfactory discrimination for mortality in our cohort despite lacking any input from patient characteristics besides the amount of inotropes received.

There are certain factors that influence survival but are not included in the existing risk scores. These include the site of ECMO cannulation and the concurrent use of IABP, fluid balance, the postoperative ejection fractions, some of which were not available for analysis in our stıdy.^23,24 There are also other significant factors such as center experience that contribute to ECMO outcomes. Different centers may have varying protocols and experience in ECMO care and inotrope management. ²⁵ In our study, institutional protocols guided by ELSO protocols were followed for hemodynamic, metabolic and respiratory management of ECMO patients. Future studies that investigate novel risk scores for PC-ECMO patients may benefit from evaluating the value of these factors for their association with survival and weaning success.

Conclusion

Current risk scores for PC-ECMO patients are not satisfactory for prediction of weaning and survival. 24th hour VIS can be a valuable parameter for risk analysis and prospective studies can investigate the incorporation of VIS in other scores to create a novel PC-ECMO risk scoring system.