Early outcomes of moderate-to-high-risk pediatric congenital cardiac surgery and predictors of extracorporeal circulatory life support requirement

Background Cardiopulmonary failure refractory to medical management after moderate-to-high-risk congenital cardiac surgery may necessitate mechanical support with veno-arterial extracorporeal membrane oxygenation (ECMO). On the extreme, ECMO can also be initiated in the setting of cardiac arrest (extracorporeal cardiopulmonary resuscitation, ECPR) unresponsive to conventional resuscitative measures. Methods This was a single-center retrospective cohort study of patients (n = 510) aged <3 years old who underwent cardiac surgery with cardiopulmonary bypass with a RACHS-1 score ≥3 between 2011 and 2014. Perioperative factors were reviewed to identify predictors of ECMO initiation and mortality in the operating room (OR) and the intensive care unit (ICU). Results A total of 510 patients with a mean surgical age of 10.0 ± 13.4 months were included. Among them, 21 (4%) patients received postoperative ECMO—12 were initiated in the OR and 9 in the ICU. ECMO cannulation was associated with cardiopulmonary bypass duration, aortopulmonary shunt, residual severe mitral regurgitation, vaso-inotropic score, and postprocedural lactate (p < 0.001). Of the 32 (6%) total deaths, 7 (22%) were ECMO patients—4 were elective OR cannulations and 3 were ICU ECPR. Prematurity [hazard ratio (HR): 2.61, p < 0.01), Norwood or Damus–Kaye–Stansel procedure (HR: 4.29, p < 0.001), postoperative left ventricular dysfunction (HR: 5.10, p = 0.01), residual severe tricuspid regurgitation (HR: 6.06, p < 0.001), and postoperative ECMO (ECPR: HR: 15.42, p < 0.001 vs. elective: HR: 5.26, p = 0.01) were associated with mortality. The two patients who were electively cannulated in the ICU survived. Discussion Although uncommon, postoperative ECMO in children after congenital cardiac surgery is associated with high mortality, especially in cases of ECPR. Patients with long cardiopulmonary bypass time, residual cardiac lesions, or increased vaso-inotropic requirement are at higher risk of receiving ECMO. Pre-emptive or early ECMO initiation before deterioration into cardiac arrest may improve survival.


Introduction
In surgical repair of moderate-or high-risk congenital heart disease (CHD), extracorporeal membrane oxygenation (ECMO) is utilized for patients who fail to wean from cardiopulmonary bypass (CPB) and experience intractable low cardiac output syndrome (LCOS), pulmonary hypertension, or sudden cardiac arrest (1,2).The in-hospital survival of pediatric patients receiving postcardiotomy ECMO ranges from 22% to 49% (3), with some studies reporting worse outcomes in those acutely cannulated during extracorporeal cardiopulmonary resuscitation (ECPR) (4,5).Our institution's data of 180 consecutive postoperative cardiac surgery ECMO cases from 1990 to 2007 identified neurological complications, renal dysfunction, prolonged ECMO duration greater than 3 days, repeat ECMO initiation, and the lack of a heart transplant exit strategy to be associated with in-hospital mortality (6).Interestingly, the indications for ECMO and ECPR were not associated with mortality in this analysis (6).Ischemic brain injury was the most common cause of death among patients who received ECPR (6).In recent years, surgical and intensive care practices have evolved, making it worthwhile to re-evaluate prognosticators of postoperative ECMO and mortality.
In this report, we summarized our institution's more recent experiences with pediatric patients after moderate-to-high-risk congenital cardiac surgery to identify the predictors of ECMO initiation and its effect on in-hospital and 1-year survival.We anticipated that ECPR is associated with poorer survival than elective ECMO.Being able to recognize the demographic, operative, echocardiographic, and biochemical predictors for ECMO could allow for early ECMO initiation and subsequently improved survival.

Patients and methods
A total of 510 consecutive children aged under 3 years who underwent surgery with CPB and RACHS-1 score ≤3 at the Hospital of Sick Children in Toronto between September 2011 and November 2014 were identified and retrospectively reviewed.The study was approved by the Research Ethics Board at the Hospital for Sick Children, and patient consent was waived.

ECMO protocol
All neonates and infants were treated according to a standardized protocol for ECMO as previously described (6,7).Indications for ECMO included the inability to separate from CPB, LCOS, cardiac arrest (institutional practice of ECPR started in 2000), and shunt thrombosis, with the decision ultimately left to the discretion of the interdisciplinary clinical team.In short, in patients within 1 week following cardiotomy, the aorta and atria were directly cannulated through an open chest for expeditious commencement of support and facilitation of effective open cardiopulmonary resuscitation as required.In older patients and those who received delayed postoperative ECMO, peripheral cannulation via the neck or femoral vessels was performed.In this study, patients were <3 years of age with body weight <20 kg; hence, neck access was used exclusively in cases involving peripheral cannulation.Left heart decompression, for left chamber dilation on echocardiography or significant pulmonary edema on chest radiography, was achieved through left atrial cannulation and drainage, or through surgical or balloon atrial septostomy (8).Pump flow was titrated between 100 and 120 ml/kg/min (150-200 ml/kg/min in aortopulmonary shunt lesion), targeting adequate end-organ perfusion and achieving normalization of arterial blood gases, systemic venous saturation, and lactic acid clearance.Anticoagulation was maintained by a continuous heparin infusion at a rate of 50 U/kg, targeting an activated clotting time of 180-200 s.Broadspectrum prophylactic antibiotics were routinely given in cases of open chests.Echocardiography was performed frequently to monitor cardiac recovery, left ventricular (LV) thrombus formation, and residual cardiac lesions and to inform discussion for a durable ventricular assist device.

Data collection
Patient demographics including sex, age, prematurity, birth weight, presence of genetic syndromes, baseline biventricular function, and congenital cardiac anatomy were collected.Operative details including CPB time, aortic cross-clamp time, deep hypothermic circulatory arrest time, and type of procedure were obtained.Ventricular function, residual valvular lesions, and other defects on early postoperative echocardiography were examined.Blood arterial lactate level, pH, oxygen saturation of arterial blood (SaO 2 ), vasoactive-inotropic score (VIS), and near-infrared spectroscopy for cerebral oximetry (NIRS) at the time of postrepair transesophageal echocardiography and on admission to intensive care unit (ICU) were documented.The setting of veno-arterial ECMO initiation [i.e., operating room (OR) or ICU] was also noted.For terminology in this study, ECMO encompassed both entities of ECPR and elective ECMO.ECPR was defined on the basis of the Extracorporeal Life Support Organization definition of circulatory support in the context of failure to achieve sustained return of spontaneous circulation using conventional cardiopulmonary resuscitation (CPR) (9).Elective ECMO was defined by mechanical circulatory support for failure to wean from CPB or inadequate support despite maximal inotropic agents.LCOS was defined as a cardiac index <2.2L/min/m 2 with signs of end-organ malperfusion in the absence of hypovolemia.Other variables such as delayed chest closure, length of intubation, ICU stay, hospital stay, and death were also analyzed.The primary outcome was in-hospital survival and predictors for ECMO initiation in intraoperative and ICU settings.The secondary outcomes were predictors of postoperative mortality in patients supported with ECMO, compared to those not requiring ECMO support.

Statistical analysis
Continuous variables were described using mean and standard deviation; categorical variables were described using frequencies and proportions.Between-group (ECMO vs. no ECMO) differences in the continuous and categorical variables were evaluated using the Wilcoxon rank-sum test and Fisher's exact test, respectively.Survival was characterized using the Kaplan-Meier method, and between-group differences were assessed using the log-rank test.The impact of receiving ECMO in ICU among those who did not receive ECMO in the OR was quantified using a multistate model, with "no ECMO in OR" being the initial state, "ECMO in ICU" being the transient state, and "death" being the terminal state.ECMO in OR is considered as a binary outcome, whereas ECMO in ICU and mortality are considered as time-toevent outcomes.Administrative censoring was applied at year 1 for the endpoint of death.Univariate analysis was performed because multivariate analysis was not feasible given the low incidence of ECMO events.All analyses were conducted assuming a significance level of 5% and implemented using R v4.0.3 (10).

Patients and operative characteristics
A total of 510 consecutive children were included, of which 21 received postoperative ECMO (Table 1) between postoperative days 0 and 6, with 14 patients cannulated on the day of surgery.The average age of operation in the ECMO group was 4.2 ± 7.8 months, compared to 10.0 ± 13.6 months in the non-ECMO group (p = 0.047), with no differences in the prevalence of prematurity.Aortic cross-clamp (140 ± 90 vs. 91 ± 55 min, p = 0.03) and CPB duration (250 ± 99 vs. 128 ± 58 min, p < 0.001) were significantly longer in the ECMO group.Aortopulmonary shunt, pulmonary arterioplasty, and coronary repair were more prevalent in the ECMO group.On average, patients had normal biventricular function at baseline in both groups.

Predictors of ECMO
Predictors of ECMO use in the OR were CPB duration, unplanned procedure, unifocalization, coronary repair, aortopulmonary shunt, residual severe MR, intraoperative lactate, and VIS immediately postrepair (Table 3).Predictors of ECMO use in the ICU were pulmonary arterioplasty, aortopulmonary shunt, residual severe tricuspid regurgitation (TR), and VIS on admission to the ICU.

Survival
In-hospital survival was 90.5% and 98.0% in the ECMO and non-ECMO groups, respectively.At a median follow-up of 10.8 (IQR: 2.6-22.6)months, survival was 66.7% (n = 14/21) in the ECMO group and 94.9% (n = 464/489) in the non-ECMO group.Of patients who received ECMO, survival at 2 and 10 years was 75.0% and 62.5% compared to 98.7% and 93.6% among patients who did not receive ECMO in the OR, respectively (Figure 2).Of the seven deaths in the ECMO group, four occurred after elective ECMO initiation intraoperatively and three after ECPR in the ICU.Of the 32 total deaths, 12 occurred in the ICU, 3 in the non-ICU setting within the hospital, and 17 in the community.The baseline characteristics of ECMO patient survivors and nonsurvivors are shown in Supplementary Table S1.

Predictors of mortality
Univariate analysis revealed that the predictors for mortality in the general cohort were prematurity, Damus-Kaye-Stansel or Norwood procedure, pulmonary vein repair, coronary repair, moderate-to-severe LV dysfunction, residual severe TR, and

Discussion Prevalence of postcardiotomy ECMO
This single-center retrospective cohort study analyzed the predictors, settings, and outcomes of ECMO initiation among 510 consecutive pediatric patients who underwent moderate-to-high-   (14).ECPR comprises 27%-39% of total ECMO cases (6, 11, 13), with ECPR rates as high as 50% described in a cohort of patients aged ≤6 weeks (12).Our observed ECMO and ECPR rates were within the upper range of the existing literature due to the inclusion criteria of RACHS-1 ≥ 3.

Biochemical predictors of ECMO initiation
In keeping with findings from other centers, arterial lactate and VIS were identified as predictors of postoperative ECMO after pediatric cardiac surgery (12).In the ECMO group, the intraoperative median VIS and arterial lactate were 16.6 (8.6-35) μg/kg/min and 7.7 (4.5-10) mmol/L, respectively.The multivariate regression analysis by Kuraim et al. identified peak VIS >30 (OR: 1.02; p < 0.001) and peak lactate >6 mmol/L (OR: 1.2; p = 0.003) as makers for ECMO initiation within 48 h postrepair (12).Conversely, Charpie et al. proposed that uptrending serial lactate levels were a more sensitive and specific indicator for the composite outcome of ECMO or early death than singular lactate levels.They found that initial lactate >6 mmol/L and increasing lactate of >0.75 mmol/L/h had positive predictive values of 38% and 100%, respectively (15).Although lactate was not trended in our study, the poor prognostication associated with lactate >6 mmol/L, often referred to in the literature, is also reflected in the findings.The median VIS score in our ECMO group was comparatively lower, possibly owing to a lower threshold for ECMO initiation before maximally tolerated vaso-inotropic agents are employed, as suggested by 9 out of 21 patients being cannulated in the OR.

Surgical predictors of ECMO initiation
Our univariate analysis identified aortopulmonary shunt, unifocalization, and coronary repair as predictors of intraoperative ECMO.Aortopulmonary shunts were often performed in conjunction with other procedures, including unifocalization, truncus arteriosus repair, pulmonary vein repair, and pulmonary arterioplasty.Aside from shunt thrombosis, the indication for ECMO in the aortopulmonary cohort can be driven by inadequate repair of other cardiac lesions.Unifocalization being a risk factor for intraoperative ECMO could be attributed to the sequelae of pulmonary edema and right ventricular (RV) dysfunction associated with major aortopulmonary collateral arteries or shunt run-off causing coronary hypoperfusion in cases of staged repair.Suboptimal coronary repair or collateral coronary injury during repair of the primary lesion poses difficulty with weaning from CPB, hence greater odds of intraoperative ECMO initiation.
Although there is a significant overlap between predictors of ECMO in the OR and in the ICU, pulmonary arterioplasty and residual severe TR were recognized as predictors of ECMO institution in the ICU, but not in OR.The severity of right heart lesions and their associated RV dysfunction are often unmasked in the ICU after perioperative hypotension and hypovolemia are treated.The association of ICU ECMO with pulmonary arterioplasty was initially hypothesized to be due to baseline RV dysfunction, but all three patients had normal functioning RV.However, residual pulmonary artery stenosis was noted in two of the three patients, with one of the patients also demonstrating additional preoperative moderate LV dysfunction.
In a meta-analysis of 16 heterogeneous cases of pediatric ECMO for CHD by Wu et al., the pooled estimate of residual lesions requiring reoperations was 14.9% (I 2 = 80.3%) (16).This percentage was lower than in a large observational study by Agarwal et al., which found hemodynamically significant residual lesions requiring reintervention in 35 (28%) of 119 postoperative pediatric cardiac surgery patients, with the predominant lesion being branch pulmonary artery stenosis (n = 7/13) (17).They reported a superior rate of decannulation (p = 0.004) and in-hospital survival (p = 0.025) when residual lesions were identified in the first 3 days of ECMO support (17).Our findings corroborated the importance of avoiding residual defects, particularly residual pulmonary artery stenosis, with certain lesions being harder to appreciate until hemodynamic optimization in the ICU.

ECMO survival and predictors of mortality
The in-hospital survival of postcardiotomy pediatric ECMO in this study was 90.5%.Historically, it was reported to be 37% at our institution (18), which is similar to the rates reported by other single-center studies of 37%-41% (17, 19).In a multicenter study involving 998 children from 37 American centers, Gupta et al. reported postoperative ECMO in-hospital survival of 48.1% (20).A meta-analysis of 43 international studies showed a pooled in-hospital mortality estimate of 56.8% (95% CI, 52.5%-61.0%,I 2 = 74.2%)after pediatric CHD surgery (16).In that same study, multivariate meta-regression revealed single-ventricle physiology and renal failure as independent risk factors for in-hospital nonsurvival (16).The results from our study demonstrated comparably better post-ECMO survival than current literature; however, the mortality of surgery in single-ventricle physiology remains notably high and accounts for four of the seven deaths in our ECMO group.
Lactate is a well-investigated marker of mortality in the ECMO population and has been used to inform appropriateness for withdrawal of care (21).Our study revealed intraoperative and end-of-surgery lactate and VIS to be predictors of mortality.Similarly, Shah et al. also described high arterial lactate at the onset of ECMO to be a predictor of non-survival (lactate 14.4 ± Kaplan-Meier survival curve of patients with ECMO initiation in the operating room (red) compared to those without ECMO initiation (blue).Shaded areas represent 95% confidence intervals.The table below the graph shows the number of patients at risk in each cohort over time.Administrative censoring was applied at year 1.HR: 7.2 (2.5-20.7),p < 0.001.(22).Prognostication is a multifaceted and dynamic process with caseby-case nuances.Whether trended or stand-alone, the lactate level is one of many markers to inform the larger clinical picture.

Location of ECMO initiation
We found similar survival rates for ECMO cannulation in the OR and ECMO cannulation in the ICU.This is consistent with historical in-hospital survival data from our institutions (40% in OR vs. 37% in the ICU or the cath lab; p = 0.62) (6).Many pediatric studies that have explored the setting of ECMO initiation in the OR vs. ICU also reported no statistically significant difference in survival between locations (3,11,23,24).Only one study demonstrated significantly better in-hospital survival when ECMO runs were initiated in the OR rather than in the ICU (64% vs. 29%, p = 0.003), partially because 56% of their ICU ECMO cohort were ECPR cases (25).Although not specific to postcardiotomy, a recent multicenter pediatric registry study of ECPR for in-hospital arrest showed an adjusted mortality odds ratio of 1.04 (95% CI, 1.01-1.07)for each 5 additional minutes between CPR and the start of ECMO (26).Our results are in line with the contemporary understanding that hemodynamic status before ECMO portends survival, rather than the environment of cannulation.

ECPR vs. ECMO
There was no statistically significant survival difference between ECPR and elective ECMO.Our previous institutional data also demonstrated no survival disadvantage of ECPR (p = 0.18); in fact, survival was 46% and 35% in ECPR and ECMO groups, respectively (6) Given the relative infrequency of ECPR, two meta-analyses on postcardiotomy pediatric ECMO attempted to answer whether ECPR carries worse in-hospital survival through subgroup analyses.The results were mixed.Pooled ECPR survival in a subgroup analysis involving 12 studies was 0.37% (95% CI, 0.29%-0.46%,I 2 = 32%, p = 0.13), with the authors concluding lower mortality in ECMO for LCOS than ECPR (14).Pooled in-hospital mortality rates of 18 studies with ≤50% incidence of ECPR and 10 studies with >50% incidence of ECPR were 58.1% and 58.9%, respectively (16).In our study, ECPR had a threefold higher mortality hazard ratio than elective ECMO but failed to reach statistical significance.Early anticipation of ECMO requirement is needed to prevent ECPR and is widely adopted despite the lack of consensus or definitive survival benefit of elective or end-of-case ECMO.

Limitations
Data were retrospectively collected and derived from a single center.The study is inherently limited in generalizability by its retrospective nature and modest sample size, which precluded competing risk analysis.Although a standardized protocol for the management of all patients on ECMO was followed, this underwent some evolution over the 9 years after data collection was completed.Multiple intensivists were involved in the care of the patients with expectable variations in practice.

Conclusion
In pediatric patients undergoing CHD repair of moderate-tosevere complexity, ECMO can be initiated for the inability to wean from CPB or as rescue therapy for patients in hemodynamic extremis in the early postoperative setting.For patients who require ECMO, mortality remains considerable.CPB time, moderate-to-high LV dysfunction, significant residual

FIGURE 1 Sankey
FIGURE 1Sankey flowchart illustrating the event trajectory of all patients throughout a median follow-up of 10.8 (IQR: 2.6-22.6)months.Numbers describe the number of patients in each status.Alive and death are terminal states.
(11) car(13) surgery; among them, 21 (4.1%) received postoperative ECMO, of which there were 10 cases of ECPR.In the contemporary era, the rate of postcardiotomy ECMO in pediatric patients and neonates with congenital heart disease ranges from 3% to 4%(11)(12)(13), with a rate of 2.5% reported in systematic analysis and meta-analysis of 26 studies by Cho et al.

TABLE 3
Univariate analysis of predictors of ECMO in the OR and in the ICU.The 95% confidence interval is shown in parentheses.Bold values denote statistical significance at the p < 0.05 level.
. Shah et al. substantiated that postoperative CPR in the ICU at the time of ECMO initiation did not result in poorer survival (p = 0.59) in the CHD population (19).The 2023 Extracorporeal Life Support Organization registry-based retrospective cohort study of 2,155 CHD neonatal ECMO cases continued to show no survival difference in ECPR vs. ECMO (p = 0.31) (27).

TABLE 4
Univariate analysis of predictors of mortality.The 95% confidence interval is shown in parentheses.Bold values denote statistical significance at the p < 0.05 level., elevated lactate, and VIS emerged as predictors of ECMO initiation.Risk factors for mortality include similar predictors to those for ECMO, with ECMO itself being a predictor of death and ECPR being particularly high-risk.Given the accentuated mortality of ECPR events in the ICU, prophylactic and early ECMO initiation in patients with multiple risk factors could be a promising strategy for further investigation.
a Measurements taken at different time points are annotated by superscript numbers: in the operating room after surgical repair.bPostoperative ICU admission.lesions