Extracorporeal Membrane Oxygenation

Abstract

The utilization of extracorporeal membrane oxygenation (ECMO) for cardiopulmonary support continues to increase globally, with > 190,000 ECMO cases reported to the international Extracorporeal Life Support Organization Registry. The present review aims to synthesize important contributions to the literature surrounding the management of mechanical ventilation, prone positioning, anticoagulation, bleeding complications, and neurologic outcomes for infants, children, and adults undergoing ECMO in 2022. Additionally, issues related to cardiac ECMO, Harlequin syndrome, and anticoagulation during ECMO will be discussed.

Introduction

Extracorporeal membrane oxygenation (ECMO) is an advanced form of temporary cardiopulmonary support used for over 5 decades with considerable evolution of equipment, indications, and application. The utilization of ECMO continues to increase globally, with > 190,000 total cases reported to the international Extracorporeal Life Support Organization (ELSO) Registry.¹ Most commonly, venoarterial ECMO (VA-ECMO) is used to accomplish cardiopulmonary support, and venovenous ECMO (VV-ECMO) is preferred for those with respiratory failure, including ARDS related to COVID-19 (> 16,000 cases).^2-5,6,7 The coming sections highlight important contributions to the literature guiding mechanical ventilation during infant, pediatric, and adult ECMO. Additionally, emerging data on prone positioning (PP), extracorporeal cardiopulmonary resuscitation (ECPR), cardiogenic shock, and anticoagulation will be discussed.

Mechanical Ventilation During ECMO

Infant and Pediatric

Severe respiratory failure continues to be the primary indication for neonatal ECMO (> 70% of cases) with common etiologies including congenital diaphragmatic hernia, meconium aspiration syndrome, and persistent pulmonary hypertension of the newborn. Moreover, VA-ECMO is routinely used to provide circulatory support for infants with congenital heart disease, ECPR, inability to separate from cardiopulmonary bypass, or as bridge to transplantation. Although ECMO has improved survival in these populations, incidence of chronic lung disease and other morbidity remains substantial in survivors. Several retrospective studies of ELSO Registry data aimed to explore potential associations of mechanical ventilation and mortality in infants and children supported with ECMO (Table 1).^8-11

Table 1.

Infant and Pediatric Mechanical Ventilation Under Extracorporeal Membrane Oxygenation

A monocentric study, comprised predominantly of infants with congenital diaphragmatic hernia (66%) under VA-ECMO, assessed the incidence and severity of chronic lung disease at 28 d following neonatal ECMO. Perez-Ortiz et al⁸ separated the cohort based on the presence/absence of chronic lung disease and found > 76% of survivors developed chronic lung disease, of which 25% were severe chronic lung disease. Infants with congenital diaphragmatic hernia developed chronic lung disease (94% vs 60%) and severe chronic lung disease (42% vs 7%) more frequently than other diagnoses (eg, persistent pulmonary hypertension of the newborn/meconium aspiration syndrome). Secondary analyses were utilized to compare the development of chronic lung disease in accordance with diagnosis. Analysis of the congenital diaphragmatic hernia group demonstrated that 64% vs 14% developed chronic lung disease, and 88% vs 11% met severe criteria, with no significant differences in perinatal variables between groups. Paradoxically, the comparison of subjects with meconium aspiration syndrome found that 38% of these infants did not develop chronic lung disease, and 62% met criteria for mild chronic lung disease. Moreover, birthweight was the sole perinatal characteristic that was predictive of chronic lung disease in this population. Other risk factors for chronic lung disease included diagnosis, lower birthweight, younger age at cannulation, and need for ECMO beyond 7 d.

Polito et al¹⁰ used the ELSO Registry to evaluate the impact of ventilator support prior to ECMO on mortality in children $\ge$ 29 d to $\le$ 18 y with pediatric ARDS.¹² Logistic regression analysis was used to examine associations between ventilatory parameters and mortality. Pre-ECMO factors including longer duration of ventilation, higher oxygenation index, and lower pH being independently predictive of mortality. Notably, median pre-ECMO PEEP was 10 (interquartile range [IQR] 7–12) cm H₂O and considerably lower −6 (IQR −10 to −3) cm H₂O than ARDSNet PEEP/F_IO2 protocol recommendations. Further, the median ΔP (peak inspiratory pressure [PIP] −PEEP) was 23 cm H₂O, higher than pediatric guideline recommendations (ΔP < 15 cm H₂O).¹³ Although, elevated ΔP and lower PEEP (compared to ARDSNet) have been associated with pediatric ARDS mortality in observational studies, neither was associated with mortality in the present study.^7,14

Similarly, Blauvelt et al⁹ examined ventilator settings 24 h post cannulation in children undergoing ECMO for respiratory failure and found PEEP > 10 cm H₂O and F_IO2 $\ge$ 0.45 and 0.6 were independently associated with mortality compared with reference groups (PEEP > 8 cm H₂O and F_IO2 < 0.40). In a secondary analysis of survivors, PEEP 8–10 cm H₂O and ΔP > 16 cm H₂O were associated with shorter ECMO duration and F_IO2 $\ge$ 0.6 with prolonged ECMO support. The association between PEEP > 10 cm H₂O and F_IO2 $\ge$ 0.45 at 24 h with mortality may be attributed to heterogeneity of underlying condition, unmeasured or residual confounding, and inability to reduce settings further in both undersupported and moribund subjects. In survivors, the application of PEEP 8–10 cm H₂O contributed to shorter ECMO support, whereas those with higher F_IO2 $\ge$ 0.6 experienced more protracted ECMO support. Similarly, in cohort of children on VV-ECMO, Friedman and colleagues¹¹ found that F_IO2 at 24 h after ECMO was the only ventilator setting independently associated with mortality, odds increasing 13% for each 0.1 change in F_IO2. Importantly, higher F_IO2 retained association with mortality after adjusted analysis, and this relationship was independent of arterial oxygen saturation (S_aO2) on ECMO, despite worse lung disease in moribund subjects prior to ECMO and lower S_aO2 during VV-ECMO. These findings suggest that early reduction of F_IO2 may contribute to improved outcomes for children on VV-ECMO and are in accordance with previous work.¹⁵

Collectively, these studies place impetus on further definition of optimal thresholds for tidal volume, breathing frequency, F_IO2, PEEP, and ΔP for infants and children during ECMO.¹⁶ Importantly, these data must be interpreted in context with limitations inherent of retrospective research and require further prospective investigation. Further, analysis of ELSO Registry data includes blood gas values obtained before and 24 h after cannulation, which limits the ability to analyze ventilator titration over time. Hence, longitudinal assessment of ECMO support, ventilator titration, and impact on gas exchange from cannulation to liberation is desirable to identify temporal interactions and modifiable factors.

Adult

Tonna and colleagues¹⁷ examined if mechanical ventilation parameters and blood gas values influenced survival in a cohort of 7,488 adults who underwent ECPR. After adjustment for clinically relevant covariates, hospital survival was reduced with higher P_aO2 on ECMO (odds ratio [OR] 0.69 per 1 SD increase [95% CI 0.64–0.74]) and any relative Δ change in P_aCO2 (up or down) between pre-arrest to 24 h. Moreover, survival worsened with PIP > 20 cm H₂O (OR 0.69 per 1 SD [95% CI 0.64–0.75]), F_IO2 > 0.40 (OR 0.75 per 1 SD [95% CI 0.69–0.82]), higher breathing frequency (OR 0.82 [95% CI 0.75–0.89]), and higher ΔP (OR 0.72 per 1 SD ∼7 cm H₂O increase [95% CI 0.65–0.79]). Collectively, these findings underscore need for meticulous adjustment of sweep gas flow and oxygen fraction along with ventilator titration to avoid large magnitude swings in P_aCO2 and P_aO2. In the absence of randomized controlled trial (RCT) data, these results support current expert recommendations to provide protective ventilation, reduce breathing frequency, and provide adequate PEEP, generally ≥ 10 cm H₂O with concomitant reduction of F_IO2 and ΔP following transition to ECMO.¹⁸ Importantly, the optimal mechanical ventilation strategy during ECMO remains elusive, and some advocate for early extubation to avoid complications from mechanical ventilation, delirium, and decreased mobility (awake ECMO). Although, awake ECMO is feasible and was not associated with complications for both neonates and adults,^19,20 further investigation is needed to guide patient selection.

Prone Positioning During VV-ECMO

Data from adult studies demonstrate that PP for (> 16 h/d) in severe ARDS can improve gas exchange and confers survival benefit when used in conjunction with a lung-protective ventilation strategy.²¹ It has been demonstrated that PP during VV-ECMO is safe and feasible, with mounting evidence that PP improves survival for severe ARDS supported with VV-ECMO. In the absence of randomized data, several observational studies, systematic reviews, and meta-analyses aimed to assess the impact of PP on survival for adults undergoing VV-ECMO for ARDS (Table 2). Papazian et al²² included 13 studies for meta-analysis with the objective of comparing VV-ECMO with PP versus VV-ECMO without PP on 28-d survival. Prone subjects had higher survival (74% vs 58%) at 28 d, and were on mechanical ventilation longer (standard mean difference 11 [95% CI 9.2–13.5] d). Congruently, an individual-subject meta-analysis pooled subjects from 5 studies investigating the associations between use of PP during VV-ECMO and ICU mortality.²³ Although the unadjusted mortality was lower in the PP group (41% vs 53%), PP was not associated with reduced ICU mortality following Cox multiple regression analysis (hazard ratio [HR] 0.67 [95% CI 0.42–1.06]). Post hoc analysis demonstrated the strongest associations between PP during ECMO and decreased mortality were observed in those $\le$ 50 y, body mass index $\le$ 30 kg/m², Sequential Organ Failure Assessment > 12, P_aO2/F_IO2 $\le$ 60, and in subjects with less exposure to mechanical ventilation.

Table 2.

Prone Versus Supine Position for ARDS Under Venovenous Extracorporeal Membrane Oxygenation

Petit and investigators^24,25 provided novel insight by defining response to PP as increase of compliance of $\ge$ 3 mL/cm H₂O after 16 h of PP. The results of the propensity-scored, matched analysis compared combination of VV-ECMO with PP to supine VV-ECMO, using the end point of time to successful weaning, with death as a competing risk. Although groups had similar ECMO durations (∼15 d) after propensity-score matching, proned subjects had a higher probability of being liberated from ECMO (0.75 vs 0.54) (OR 1.54 [95% CI 1.05–2.58]) and alive at 90 d (20% vs 42%), with longer ICU stay (37% vs 27%) compared to the supine group. Moreover, responders (53%) demonstrated static compliance gains, ∼6 mL/cm H₂O versus 0 mL/cm H₂O, P < .01 for non-responders.

Building upon these findings, a secondary analysis of the EuroPronECMO study²³ explored the temporal associations of PP, with the primary end point of being alive at discharge from ICU at 90 d and improved respiratory system compliance (C_RS).²⁶ Time until PP during ECMO was analyzed as a continuous variable and subsequently explored as a binary outcome: early PP ≤ 5 d versus late PP > 5 d after ECMO onset. Importantly, late PP was associated with lower probability of being discharged alive (61% early vs 36% late), and this benefit remained after adjustment (HR 2.52 [95% CI 1.66–3.81]). Furthermore, the early group demonstrated improved C_RS (4 mL/cm H₂O vs 0 mL/cm H₂O), which was associated with shorter ECMO duration and more rapid weaning at 60 d. Measurement of C_RS as a response variable is less likely to be confounded by ECMO support (compared with oxygenation metrics) and provides objective response criteria to PP during VV-ECMO. Additionally, this study was the first to estimate a temporal window of efficacy for PP and suggest benefit for earlier institution. Finally, prompt identification of non-responders (no improvement) may be used to inform ECMO guidelines and reduce unnecessary, labor-intensive tasks in a climate of capacity issues and strained resources.

The use of PP has increased throughout the pandemic, with large multi-center cohorts reporting > 60% utilization in adults with COVID-19–related severe ARDS.^27,28 A global multi-center by Zaaqoq et al²⁹ reported 31% utilization of PP during VV-ECMO and confirmed favorable survival benefit (HR 0.31 [95% CI 0.14–0.68]). Similarly, Massart et al³⁰ found PP before and after VV-ECMO initiation were independently associated with lower mortality following multivariable analysis. Among propensity score–matched subjects alive at decannulation, PP had lower mortality and had longer ECMO duration (15 d vs 10 d) (95% CI 5–16) and more days alive after cannulation (90 d vs 32 d) (95% CI 5–16) compared with their supine counterparts. Hence, PP during VV-ECMO appears to confer survival benefits in select ARDS phenotypes; however, RCTs are needed prior to definitive conclusion that PP be routinely combined with VV-ECMO for all patients with severe ARDS.

Transfusion Threshold for VV-ECMO

In the absence of guidelines recommending a specific transfusion threshold for VV-ECMO, many centers have targeted normal hemoglobin (Hb) levels (> 12 g/dL) to maximize O₂ carrying capacity and oxygen delivery. However, associations between packed red blood cell transfusion volume and mortality have underscored the importance of defining a critical threshold for transfusion.³¹ A recent multi-center, prospective study included > 600 adults with ARDS supported with VV-ECMO and longitudinally monitored Hb levels over 28 d. Mean Hb was 10.9 g/dL before and 9 g/dL after cannulation, with 83% of subjects receiving at least one unit packed red blood cells.³² The time-dependent Cox proportional hazards analysis identified transfusion for Hb < 7 g/dL as the only cutoff associated with reduced mortality (OR 2.99 [95% CI 1.95–4.60]). Moreover, positive fluid balance at all measured time points was independently associated with transfusion rate and risk of death. These data support restrictive transfusion practices for adults with ARDS under VV-ECMO and warrant prospective interventional studies to confirm findings.

Cardiac ECMO

ECPR

The application of ECMO as an extension of cardiopulmonary resuscitation (CPR) (ECPR) is increasing in children and adults,^1,33 with contemporary survival rates of ∼20–44%.^2,17,34 Sood et al³⁴ included 30 studies for meta-analysis and assessed 30 covariates as predictors of survival in children following ECPR. Survival to discharge was 44% (95% CI 40–47), with lower lactate, higher pH, and P_aO2 before cannulation having the strongest associations with survival. These findings were corroborated by Gutierrez and colleagues,³⁵ who examined pre-cannulation factors related to VA-ECMO timing in children with myocarditis, of which 54% received ECPR. Survivors (72%) had shorter intubation-to-cannulation times (3 h vs 6 h) and demonstrated higher pre-ECMO P_aO2 (76 mm Hg vs 59 mm Hg) and pH (7.3 vs 7.2), with pre-ECMO pH retaining significance on multivariable survival improving (per 0.1) increase in pH (OR 1.18 [95% CI 1.05–1.31]). Survival of children experiencing cardiac arrest within 24 h of ECMO was lower (68% vs 76%) but not statistically significant. Importantly, children with acute myocarditis have more favorable survival (70–80%)³⁶ compared with other congenital or acquired heart disease undergoing ECPR (31–61%).³⁷ However, the association of higher pre-ECMO pH with survival is generalizable to other populations.

Non-randomized studies have demonstrated an association between ECPR and improved survival in select subject populations. However, both survival and functional neurologic recovery following ECPR for refractory outside-of-hospital cardiac arrest are poor. A recent meta-analysis compared the efficacy of ECPR versus conventional CPR for adults suffering outside-of-hospital cardiac arrest, using the primary outcome of survival with favorable neurological outcome. Pooled analysis of these data suggests that ECPR-treated subjects had higher rates of survival (22% vs 17%) and survival with favorable neurologic outcome (14% vs 8%) (OR 2.11 [95% CI 1.41–3.15]), with a number-needed-to-treat of 16, compared to conventional CPR.³⁸ Notably, both RCTs^39,40 included for meta-analysis had modest sample sizes and were terminated early for superiority benefit within the intervention group (ECPR), a characteristic known to alter the effect size during meta-analysis.⁴¹

More recently, the multi-center INCEPTION trial compared the same primary outcome, survival with favorable neurologic outcome, defined as cerebral performance category 1–2 (normal or disabled but independent at 30 d), and found no difference (20% vs 16%) for ECPR and CPR, respectively.⁴² As such, currently available evidence does not support the routine use of ECPR in patients suffering from outside-of-hospital cardiac arrest. Whereas these mixed results do not preclude ECPR from being effective in certain cases, it is plausible that larger sample sizes are needed to detect any meaningful benefit associated with ECPR. Perhaps some of these questions will be addressed by the ON-SCENE study, which aims to evaluate if more proximate ECPR deployment (on scene of event via helicopter emergency medical services teams) improves survival and costs/quality-adjusted life year in younger subjects (18–50 y) with outside-of-hospital cardiac arrest.⁴³

To date, no RCTs have assessed survival benefit of ECPR for in-hospital cardiac arrest, and observational studies have reported 20–40% survival for ECPR-treated adults who sustained in-hospital cardiac arrest.^44,45 To better understand the variation in survival benefit, Tonna and investigators⁴⁶ developed and externally validated a multivariable model, the RESCUE-IHCA score. Factors associated with mortality included age, time of day, initial rhythm, renal insufficiency, patient type (cardiac, medical, or surgical), and duration of the arrest/resuscitation. These data allow for refinement of predictive modeling and prognostication tools, which can help clinicians estimate mortality risk with moderate certainty (∼70%) and may assist in determining ECMO candidacy.⁴⁷

ECPR is complicated by intrinsic risks for bleeding, thrombosis, poor neurologic outcomes, multisystem organ failure and caveats including change in transplant status, inability to separate from ECMO. Carlson and colleagues⁴⁸ evaluated the cause of death in > 400 adults following ECPR for refractory cardiac arrest to describe those who underwent discontinuance of ECMO. Subjects were stratified according to the binary outcome of early withdrawal of life-sustaining therapy < 72 h versus $\ge$ 72 h from cannulation, using previously reported criteria.^49,50 Upward of 55% had ECMO discontinued within 72 h, most commonly related to poor neurologic prognosis, multisystem organ failure, and/or medical futility. Factors including lower pH (OR −3.1 [95% CI 2.18–2.80]), S_aO2 (OR 1.12 [95% CI 1.01–1.23]), ECMO flow (OR 7.01 [95% CI 1.47–34.00]), and higher PIP (OR 0.84 [95% CI 0.71–1.00]) were independently associated with early withdrawal of life-sustaining therapy following ECPR.

Cardiogenic Shock

Refractory cardiogenic shock is characterized by left ventricle (LV) dysfunction/failure, accompanied by elevated pulmonary venous pressure and pulmonary edema, resulting in reduced respiratory compliance, impaired gas exchange, and high risk for ventilator-induced lung injury.⁵¹ The use of VA-ECMO for adults with cardiogenic shock remains controversial, with reported survival estimates 58–62%^52,53 and 44% on ECMO mortality.⁵³ Common complications included renal failure (40–51%), bleeding (28–49%), and multisystem organ failure (37%), whereas age < 60 y, shorter ECMO duration, and presence of infection were associated with reduced in-hospital mortality.⁵³ Bertic et al⁵⁴ included 92 studies (n = 6,836) for meta-analysis to evaluate factors associated with short-term survival and favorable neurologic outcome in adults treated with ECPR (44% in-hospital cardiac arrest). Pooled estimates of survival and favorable neurologic outcome across all studies were 25% (95% CI 22–28) and 16% (95% CI 13–19), respectively. Meta-regression of these data demonstrated lower lactate, presence of a shockable rhythm, shorter CPR duration, and higher baseline pH were associated with short-term survival.

The ECMO in the Therapy of Cardiogenic Shock trial randomized subjects meeting criteria for rapidly deteriorating cardiogenic shock to receive early VA-ECMO or conservative therapy with retained ability for VA-ECMO cannulation. The primary end point, all-cause mortality at 30 d, was similar between groups (50.0% vs 47.5%) (HR 1.10 [95% CI 0.66–1.87]), and nearly 40% of the conservative group ultimately underwent VA-ECMO. Adverse events included bleeding (30%), leg ischemia (14%), stroke (3%), with the occurrence of pneumonia (31%) and sepsis (∼40%) being similar between groups.⁵⁵ Although early ECMO was not associated with mortality benefit in subjects with rapidly deteriorating or severe cardiogenic shock, nearly 40% of the conservative group (no early ECMO) were ultimately cannulated for hemodynamic instability, defined as an increase in serum lactate ≥ 3 mmol/L within 24 h. These findings suggest the benefits of early VA-ECMO are not superior to usual care and highlight the importance of establishing a threshold for VA-ECMO initiation. Currently, several ongoing RCTs aim to determine if VA-ECMO^56,57 or VA-ECMO with concurrent LV unloading via intra-aortic balloon pump⁵⁸ or Impella (Abiomed, Danvers, Massachusetts)⁵⁹ provides superior outcomes compared to usual care.

Harlequin Syndrome

Harlequin syndrome also called North-South syndrome⁶⁰ or watershed phenomena⁶¹ is a known caveat of VA-ECMO specific to patients with improved cardiac output and unresolved lung injury. Harlequin syndrome is characterized by the development of cerebral and coronary hypoxemia as the LV ejects deoxygenated blood from the native pulmonary circulation, whereas the lower portion of the body is perfused by the ECMO circuit resulting in regional differences in P_aO2.^62-66 Germaine to this concept is understanding that dual circulations exist and interact during VA-ECMO. For example, when utilizing a peripheral femoral (femoral vein-femoral artery) cannulation strategy, blood exits the oxygenator and perfuses retrograde via the femoral artery toward the proximal aortic arch and coronary arteries. Paradoxically, with improved LV function, hypoxemic blood from the native pulmonary circulation is ejected in opposition to oxygenated ECMO flow. The mixing zone or interface between circulations is variable in proximity within the aorta relative to pressure and flow differences. This relationship has been explored via in vitro computational fluid dynamics assessment of interactions between native and extracorporeal circulations using cadaver⁶⁰ or life-sized vascular models.⁶¹ These simulations demonstrate that with residual cardiac index $\le$ 1 L/min/m² during constant ECMO flow (3–4 L/min) the mixing zone shifted distally toward the abdominal aorta. Moreover, the inability to shift the mixing zone proximally with increased ECMO flow suggests that myocardial hypoxia could develop before cerebral hypoxia, raising concern for secondary end-organ injury.

In contrast to adults or larger children, infants are cannulated via right common carotid artery (A_car) and jugular vein (V_j) for V_j-A_car ECMO. As highlighted by Levy et al,⁶⁷ even with the arterial cannula in closer proximity to the aortic arch, there is potential for unrecognized Harlequin syndrome in infants with moderate-severe LV dysfunction and concomitant respiratory failure. As cardiac function recovers, coronary hypoxemia may occur when the mixing zone is below the brachiocephalic trunk (Fig. 1) with otherwise acceptable peripheral oxygenation. If Harlequin syndrome is suspected based on abrupt clinical, laboratory, or electrocardiogram changes, a multimodal assessment of overall cardiopulmonary function is warranted (echocardiography, chest radiograph, near-infrared spectroscopy). Management of Harlequin syndrome hinges upon both left and right ventricle function and can include conversion to V-AV ECMO, central cannulation, or VV-ECMO for patients with sufficient cardiac output. Additional supportive measures include sedation and neuromuscular blockade to reduce oxygen consumption and PP to optimize pulmonary mechanics and native oxygenation.

Fig. 1.

Hidden Harlequin syndrome in infants and children with left ventricle (LV) dysfunction and severe respiratory distress syndrome under venoarterial extracorporeal membrane oxygenation (V-A ECMO). Asterisk denotes mixing point according to LV function. For severe (A) and moderate (B) LV dysfunction, the mixing zone is below the brachiocephalic trunk, and Harlequin syndrome may go undetected as peripheral S_pO2 is adequate; however, this scenario may result in coronary ischemia and secondary injury. (C) In context with minimal LV function, V-A ECMO provides full circulatory support and sufficient oxygenation coronary arteries. (D) As cardiac function improves and LV output increases, the mixing zone shifts past the brachiocephalic trunk and is clinically detectable as reduced S_pO2 in the lower limbs, with normal S_pO2 via the right arm as this reflects mostly oxygenated blood streaming from the right common carotid arterial cannula. From Reference 67, with permission.

Acute Brain Injury

Following transition to ECMO support, the exposure to rapid P_aCO2 reduction likely attributes to acute brain injury via dose-dependent vasoconstriction and cerebral ischemia. Joram et al⁶⁸ investigated the relationship between P_aCO2 change after ECMO initiation on the composite of acute neurologic event in a retrospective analysis of 3,583 infants supported with ECMO for respiratory failure. The P_aCO2 values prior to ECMO initiation and 24 h after were used to calculate relative change in P_aCO2, denoted by equation: relative ΔP_aCO2 = [24 h P_aCO2 − pre-ECMO P_aCO2]/pre-ECMO P_aCO2.

The primary outcome, acute neurologic event at 28 d, occurred in 17%: 11% cerebral hemorrhage, 6% seizures, 3% ischemic stroke, and 6 brain deaths. Overall mortality was ∼21% and higher among infants with acute neurologic event compared to those without (45% vs 16%, P < .001). Median relative ΔP_aCO2 was −29.9% (IQR −46.2 to −8.5), and subjects with ΔP_aCO2 reduction > 50% were more likely to suffer acute neurologic event (ANE) (OR 1.78 [95% CI 1.31–2.42]), an effect that persisted after adjustment for clinically relevant confounding factors (OR 1.94 [95% CI 1.29–2.92]).

For the secondary analysis, subjects were categorized according to calculated relative ΔP_aCO2 values to explore the association of ΔP_aCO2 reduction > 50% and risk for acute neurologic event/death within subgroups, using minimal relative ΔP_aCO2 0 ± 10 mm Hg as a reference. Logistic regression analysis identified age, prematurity, pre-ECMO cardiac arrest, bicarbonate level < 12 mmol/L, and pre-ECMO blood pressure as independent risk factors for acute neurologic event/death, whereas bicarbonate > 27 mmol/L, meconium aspiration syndrome diagnosis, and VV-ECMO mode were protective from the primary outcome. Importantly, meconium aspiration syndrome as a diagnosis was found to be protective of acute neurologic event/death. However, the secondary analysis demonstrates that when infants with meconium aspiration syndrome were exposed to relative ΔP_aCO2 reduction > 50% this group demonstrated a stronger association with acute neurologic event/death (OR 3.76 [95% CI 1.35–10.49]) compared with the entire cohort (OR 1.94 [95% CI 1.59–2.92]) and those with congenital diaphragmatic hernia (OR 1.54 0.84–2.81]). Further, an association was found between all relative ΔP_aCO2 category (increase or decrease) and 28-d mortality as compared with minimal change (Fig. 2). Although the study design does not allow for causal inferences to be made, these data are reproducible^69,70 and support that P_aCO2 is a modifiable risk factor, thus warranting cautious selection of initial sweep gas and careful monitoring and titration to minimize abrupt ΔP_aCO2 correction.

Fig. 2.

Time-to-death multivariable survival analysis adjusted for all clinically relevant variables according to the relative change of P_aCO2 (Rel ΔP_aCO2) using Cox proportional hazard model. Bins of Rel ΔP_aCO2 were compared to reference group (Rel ΔP_aCO2 −10 to +10 mm Hg). Results are presented as hazard ratio and 95% CI. Rel ΔP_aCO2 = ([On extracorporeal membrane oxygenation [ECMO] 24 h P_aCO2 − Pre-ECMO P_aCO2]/Pre-ECMO P_aCO2). From Reference 68, with permission. ECMO = extracorporeal membrane oxygenation; HR = hazard ratio.

Shou and colleagues⁷¹ assessed the impact of P_aO2 values pre/post ECMO cannulation in adult subjects who underwent ECPR, exploring the outcome of composite acute brain injury occurrence, defined as ischemic stroke, intracranial hemorrhage, brain death, and seizures. Subjects were stratified according to P_aO2 level: hypoxemia (< 60 mm Hg); normoxemia (60–119 mm Hg); and mild (120–199 mm Hg), moderate (200–299 mm Hg), and severe hyperoxemia (≥ 300 mm Hg). Despite increased ECPR frequency during the 10-y period, cumulative incidence of acute brain injury decreased during the study period. In total, 16% experienced acute brain injury: 7% ischemic stroke, 6% brain death, 3% intracranial hemorrhage, and 2% seizures, with multivariable analysis finding moderate (OR 1.42 [95% CI 1.02–1.97]) and severe (OR 1.59 [95% CI 1.20–2.10]) hyperoxia under ECMO were independently associated with acute brain injury. As such, hyperoxia following ECPR likely contributes to acute brain injury and represents a modifiable factor as described in interim ELSO recommendations to maintain a safe zone during VA-ECMO and limit exposure to hypoxia or hyperoxia by adjusting the sweep gas oxygen fraction to achieve a post-oxygenator right radial P_aO2 of ∼150 mm Hg.⁷²

Management of Anticoagulation

Systemic anticoagulation is required during ECMO to prevent thrombosis yet involves a precarious balance between intrinsic risk for bleeding and thrombosis (patient, cannulas, circuit). Unfractionated heparin is the primary agent used for anticoagulation during ECMO and may be limited by unpredictable dose response, antithrombin deficiency, reliability of monitoring, or concerns for heparin resistance and heparin-induced thrombocytopenia.^73-75 The emergence of direct thrombin inhibitors including bivalirudin and argatroban as alternatives to unfractionated heparin has generated controversy and clinical equipoise on which has best efficacy and lowest risk profile.⁷⁶ To address this knowledge gap, 7 respective meta-analyses compared mortality and occurrence of major complications, including bleeding and thrombotic events between unfractionated heparin and direct thrombin inhibitors. Anticoagulation with bivalirudin was associated with reduced mortality,^77-81 less major bleeding,^74,77-80 and fewer occurrences of patient or circuit thrombosis.

Huang et al⁷⁴ included 9 retrospective studies and found that bivalirudin significantly decreased the incidence of bleeding events compared to the heparin group (relative risk 0.48 [95% CI 0.25–0.95]). In subgroup analysis according to age, bivalirudin was associated with decreased bleeding rate in children (relative risk 0.26 [95% CI 0.12–0.56]) compared with no significant difference for adults (relative risk 0.72 [95% CI 0.27–1.92). Correspondingly, Ma et al⁷⁸ found bivalirudin significantly reduced incidence of major bleeding in children (OR 0.17 [95% CI 0.04–0.66]) but not in adults (OR 0.87 [95% CI 0.46–1.62]). Furthermore, the largest analysis of 18 retrospective studies reported on major bleeding events with direct thrombin inhibitors (bivalirudin 12, argatroban 3) compared to heparin and found direct thrombin inhibitors were associated with fewer bleeding events (OR 0.48 [95% CI 0.29–0.81]). In the subgroup analysis, bivalirudin was associated with a significant reduction of major bleeding events for children (OR 0.22 [95% CI 0.13–0.38]) but not adults (OR 0.74 [95% CI 0.38–1.41]).⁸⁰ Additionally, direct thrombin inhibitors led to less pump-related thrombosis (OR 0.55 [95% CI 0.40–0.76]), and this effect was related more to comparisons between bivalirudin and heparin rather than argatroban. Patient-related thrombosis was lower with bivalirudin (OR 0.55 [95% CI 0.38–0.81]) compared with argatroban (OR 1.79 [95% CI 0.92–3.50]), and direct thrombin inhibitor–treated subjects were within therapeutic range more often (standardized mean difference 0.54 [95% CI 0.14–0.94]). Concordantly, Ma et al⁷⁸ found incidences of subject thrombosis (OR 0.58 [95% CI 0.37–0.93]) and circuit thrombosis/interventions (OR 0.40 [95% CI 0.24–0.68]) were significantly reduced with bivalirudin. Last, Wieruszewski et al⁸¹ found that heparin-treated adults were twice as likely to experience circuit-related thrombosis (OR 2.05 [95% CI 1.25–3.37]) and more likely to die (OR 1.62 [95% CI 1.19–3.37]). Although subject to the limitations of meta-analysis, bivalirudin was reproducibly associated with reduced mortality, bleeding, and thrombosis risk for children. Nevertheless, bivalirudin is a viable option for heparin-induced thrombocytopenia or heparin resistance and does not pose greater risk for adult patients supported on ECMO compared to unfractionated heparin. Regardless of anticoagulant choice, the ability to monitor and approach for adjusting anticoagulation are major determinants of hemostasis during ECMO.

Careful balance of bleeding risk and thrombotic complications during ECMO is paramount, and vigilant monitoring of coagulation parameters is requisite to titrate dosing within therapeutic range. Common laboratory tests to monitor anticoagulation include partial thromboplastin time (PTT), activated clotting time, anti-factor Xa assay, and thromboelastography (TEG).

Yabrodi and colleagues⁸² assessed the correlation between measures of anticoagulation in 100 children under ECMO following congenital heart surgery, of which 38% were ECPR. These authors found low correlation between individual anticoagulation test and low correlation between unfractionated heparin dose and anti-Xa level. Moreover, using TEG with heparinase revealed covert coagulopathy (undetected by standard laboratory parameters) in 25%, which was mainly pro-hemorrhagic, with low values for clot angle and maximum amplitude. TEG measurements can help clinicians better understand the cause of coagulopathy, target specific treatments to resolve coagulopathy, and limit unnecessary blood product administration. Importantly, standard laboratory measures of anticoagulation require blood volume (≥ 200 µL, plus waste) and vary in time to result, which may contribute to iatrogenic anemia and delay treatment.⁸³ Frydman et al describe a novel, point-of-care test, the clotting time score, able to detect pathologic clotting events with 100% sensitivity and 82% specificity as compared to PTT (0% sensitivity or specificity). Moreover, the clotting time score detected subtherapeutic anticoagulation in subjects deemed to be resistant to unfractionated heparin that was converted to bivalirudin and was able to rapidly (< 10 min) identify patients at high risk and using low blood volume (< 50 µL), which are desirable characteristics for clinical use. These data suggest that routine TEG monitoring during ECMO is beneficial and warrants further prospective investigation.

Summary

ECMO provides an opportunity to address underlying etiology of cardiopulmonary failure and reduce mechanical ventilation settings; however, careful monitoring is required to avoid rapid P_aCO2 changes and prolonged exposure to hyperoxia as both are strongly associated with negative neurologic sequalae. Emerging data suggest the combination of PP and VV-ECMO may confer a survival benefit for adults with severe ARDS; however, RCTs are needed to confirm efficacy in adults and children. The PROTECMO study identified transfusion Hb < 7 g/dL as the only cutoff associated with reduced mortality during VV-ECMO for ARDS, which supports restrictive transfusion practices for adults with ARDS. Although, unfractionated heparin is most used for anticoagulation during ECMO, emerging data demonstrate bivalirudin is associated with reduction of mortality, bleeding, thrombosis, and remains a viable alternative for patients affected by heparin-induced thrombocytopenia or heparin resistance.