The outcomes of venoarterial extracorporeal membrane oxygenation (V-A ECMO) have improved with its technological improvements and our understanding of the artificial organ. However, there are no recommended guidelines for monitoring parameters during clinical treatment using V-A ECMO. For example, macro-hemodynamic parameters (blood pressure, pulse pressure, and pulse rate) do not necessarily reflect the micro-hemodynamic status, especially the artificial flow created during ECMO, which affects the measured values. Although Bottiroli et al.,1 Li et al.,2 and Park et al.3 have reported that monitoring lactate levels is useful during ECMO management, the required testing is invasive and cannot be monitored continuously. There have been recent reports regarding the usefulness of regional cerebral oxygen saturation (rScO2) for monitoring cases of adult cardiac surgery, noncardiac surgery, sepsis, and resuscitation after cardiac arrest.4,5 Therefore, the current study aimed to evaluate whether rScO2 could be a monitoring parameter for mortality during VA-ECMO compared with mean arterial pressure (MAP) and lactate.
Study Design and Population
This retrospective observational study’s protocol was approved by the institutional review board of the Hallym University Sacred Heart Hospital, and the requirement for informed consent was waived. We searched our ECMO database for adult patients who underwent V-A ECMO for acute circulatory failure between April 2015 and October 2016. Among the 69 patients who underwent V-A ECMO, 21 patients had available cerebral oximeter monitoring data and were included in the current study. The primary outcome of interest was defined as the ability of rScO2 to discriminate between survivors and nonsurvivors at 28 days.
Measurement of Regional Cerebral Oxygen Saturation
Cerebral oxygen saturation had been measured using a near-infrared spectrometer (INVOS 5100C; Covidien, Boulder, CO) that was attached to both sides of the patient’s forehead. Each sensor uses infrared wavelengths (730 mm and 810 nm) to provide real-time data regarding the average oxygen saturation values (10% arterioles, 20% capillaries, and 70% venules) in brain tissue up to a depth of 3 cm. The rScO2 sensors had been applied within 1 h after starting the V-A ECMO, and the rScO2 data were continuously recorded.
Indications and Techniques for the Venoarterial Extracorporeal Membrane Oxygenation
The ECMO was implemented based on the decision of our institution’s critical care team. At our institution, V-A ECMO is indicated for acute circulatory failure with 1) a systolic blood pressure of <80 mm·Hg despite adequate intravascular volume replacement and the infusion of a high-dose vasopressor (norepinephrine >0.5 µg/kg/min) or 2) cardiac arrest that lasted for >10 min despite cardiopulmonary resuscitation. Patients were excluded if they had ongoing intracranial hemorrhage, terminal malignancy, loss of ability to independently perform their activities of daily living, unwitnessed cardiac arrest, or were >70 years old.
We routinely implant ECMO in a cardiac catheterization room using angiography for safety. There can be numerous technical errors when using the blind approach, despite using ultrasound, because of issues such as placing the device in the wrong direction, kinking of guidewire, and tortuous vessels. We used cannulas that were as small as possible (21 Fr for drainage and 17 Fr for infusion cannulation in V-A ECMO), and limb ischemia therefore rarely happened.
Data were collected regarding the patients’ baseline characteristics (sex, age, underlying disease), causes (extracorporeal cardiopulmonary resuscitation, cardiogenic shock, pulmonary thromboembolism, septic shock, intoxication), vital signs (MAP, pulse pressure), arterial blood gas analysis (ABGA) findings (pH, base excess, lactate levels), use of continuous renal replacement therapy, ECMO variables (anticoagulation, blood flow, duration, weaning), Swan-Ganz catheter parameters (central venous pressure [CVP], and mixed venous oxygen saturation [SvO2]), transthoracic echocardiography (TTE) parameters (ejection fraction [EF], and native cardiac output [CO]), cause of death, and outcomes (intensive care unit length of stay [LOS], hospital LOS, 28 day mortality, and neurological outcome).
During the first 7 days after the start of V-A ECMO, the EF and CO measurements were performed once per day using TTE, and the ABGA tests were performed six times per day. Mean arterial pressure was measured using an arterial cannula in a radial or femoral artery, and rScO2 was continuously monitored and recorded using a cerebral oximeter. All measured parameters were calculated as the mean value per day. The systemic vascular resistance (SVR) calculations were performed using this equation: SVR intravenous = 80 × (MAP − CVP)/(native CO measured using TTE + ECMO flow).
Continuous variables were reported as mean ± standard deviation, and categorical variables were reported as number (%). Continuous variables were analyzed using a Student’s t-test, and categorical variables were analyzed using Pearson’s χ2 test or Fisher’s exact test. For the evaluation of rScO2 as a monitoring parameter for 28 day mortality, firstly, the hemodynamic and laboratory parameters including rScO2 were compared between survivor and nonsurvivor groups during the first 7 days. Secondly, the rScO2, MAP, and lactate were evaluated using receiver operating characteristic (ROC) curves and logistic regression analysis. Thirdly, the cutoff values of rScO2 were validated using Kaplan–Meier analysis. All tests were two-tailed and p values of <0.05 were considered statistically significant. All statistical analyses were performed using IBM SPSS software (version 21.0; IBM Corp., Armonk, NY).
Demographic and Clinical Characteristic, and Initial Hemodynamic and Laboratory Parameters
Based on their survival to 28 days, the patients were categorized into a survivor group (12 patients, 57.1%) and a nonsurvivor group (9 patients, 42.9%). The patients’ demographic and clinical characteristics are summarized in Table 1. The nonsurvivor group had nonsignificantly higher frequencies of male sex (100% vs. 66.7%), coronary artery disease (16.7% vs. 11.1%), cardiogenic shock (33.3% vs. 0.0%), drug intoxication (8.3% vs. 0.0%), nafamostat mesilate treatment (100.0% vs. 88.9%), and continuous renal replacement therapy (100.0% vs. 88.9%). The two groups did not have significantly different ECMO durations (257.1 vs. 224.0 h, p = 0.67), although significant differences were observed in terms of successful ECMO weaning (91.7% vs. 11.1%), intensive care unit LOS (27.9 vs. 10.8 h), hospital LOS (39.0 vs. 10.8 days), and good neurological outcome (66.7% vs. 0.0%) (all, p < 0.05). Among the initial hemodynamic and laboratory parameters, only right and left rScO2 and lactate showed a statistically significant difference as shown in Table 2.
Comparing the Hemodynamic and Laboratory Parameters
Figures 1, 2, and 3 show the two groups’ hemodynamic and laboratory parameters. During the first 7 days, the survivor group had higher rScO2 values and lower lactate levels, compared with the nonsurvivor group (p < 0.05). Significant differences in MAP, ECMO flow, SVR, SvO2, and base excess were only observed on some of the first 7 days. No significant differences in native CO and pH were observed during the first 7 days. The difference between the rScO2 values of the survivor and nonsurvivor groups was more pronounced in cases of septic shock, compared with cases of extracorporeal cardiopulmonary resuscitation (Figure 4).
Ability of Regional Cerebral Oxygen Saturation to Discriminate Between Survivors and Nonsurvivors at 28 Days Compared with Mean Arterial Pressure and Lactate
Figure 5 shows the ROC curves for using rScO2, MAP, and lactate to discriminate between the survivor and nonsurvivor groups. The area under the ROC curves was 0.86 (95% confidence interval [CI]: 0.80–0.94, p < 0.001) for right rScO2, 0.87 (95% CI: 0.80–0.94, p < 0.001) for left rScO2, 0.91 (95% CI: 0.86–0.96, p < 0.001) for lactate, and 0.774 (95% CI: 0.69–0.86, p < 0.001) for MAP. The optimal cutoff values for right and left rScO2 and lactate were 58% (sensitivity: 78.4%, specificity: 83.3%) and 57% (sensitivity: 94.6%, specificity: 69.0%), and 2.7 (sensitivity 84.1%, specificity 86.8%), respectively. A univariate logistic regression analysis revealed that MAP, right and left rScO2, and lactate showed significant statistical differences (all, p < 0.001). Lactate and rScO2 in particular showed a more pronounced statistical difference when compared with MAP in the multivariate analysis (both, p < 0.001 vs. p = 0.018) in Table 3. Figure 6 shows that the risk of 28 day mortality was higher among patients with a right rScO2 of <58% and a left rScO2 of <57%, compared with patients with a right rScO2 of ≥58% and a left rScO2 of ≥57% (both, p < 0.001).
It is difficult to evaluate patients undergoing V-A ECMO using the conventional hemodynamic method, as their physiology is different from that of other patients with shock. Thus, the main strength of the current study is that we evaluated and compared various hemodynamic parameters, including rScO2, during the first 7 days after starting ECMO. Based on our findings, we suggest that rScO2 may be a useful monitoring parameter for 28 day mortality among patients undergoing V-A ECMO.
Conventional Hemodynamic Parameters During Venoarterial Extracorporeal Membrane Oxygenation
The hemodynamic and laboratory parameters during the first 7 days were compared between the survivor and nonsurvivor groups, although only rScO2 and lactate levels were significantly different for all 7 days (all, p <0.05). In this context, lactate levels are considered reliable for monitoring tissue perfusion and predicting mortality during V-A ECMO.1–3 For example, Park et al.3 have reported that the appropriate cutoff values for predicting mortality are 7.05 mmol/L at 6 h, 4.95 mmol/L at 12 h, and 4.15 mmol/L at 24 h. Similarly, our results revealed that the survivors had mean lactate levels of 3.85 mmol/L after the first day (vs. 10.69 mmol/L among nonsurvivors), and that the optimal cutoff value was 4.66 mmol/L (sensitivity: 75%, specificity: 75%, data not shown). Nevertheless, monitoring lactate levels remains limited because it cannot be performed as a continuous real-time measurement. In the ROC curves and logistic regression analysis, MAP also showed discriminative ability for 28 day mortality. However, in patients receiving V-A ECMO treatment, we suggest that MAP clinically could not be used alone as a monitoring parameter because the effect of antidromic artificial CO (ECMO flow) should be considered differently from the general physiology in which MAP is determined solely by native CO and SVR. Statistically, MAP was less useful than lactate and rScO2 in the current study.
In the current study, we obtained the SVR from the following equation: SVR = 80 × (MAP − CVP)/(native CO measured using TTE + ECMO flow. However, the equation for obtaining the exact SVR in V-A ECMO patients has not been established. Additional research is needed to determine the most accurate method for measuring SVR. In terms of native heart recovery, EF was significantly correlated with pulse pressure during the V-A ECMO (Pearson correlation coefficient: 0.677, p < 0.001, data not shown). However, weaning failure can lead to significant complications and death in clinical practice, so we recommend evaluating heart function as often as possible using TTE.
Interpreting and Applying Regional Oxygenation Saturation During Venoarterial Extracorporeal Membrane Oxygenation
Compared with ABGA, rScO2 monitoring is relatively simple, continuous, noninvasive, and creates a lesser burden, which makes it a potentially valuable tool for optimizing critical care. During a state of shock, the autoregulation of blood flow in the brain is relatively preserved, compared with in other organs, which makes rScO2 a trailing indicator of systemic perfusion. Thus, a low rScO2 value indicates that the patient’s condition is extremely critical, and bilaterally low rScO2 values may reflect a decreased ECMO flow, decreased SVR, native heart dysfunction, decreased intravascular volume, or Harlequin syndrome. Unilateral reductions may reflect unilateral brain complications (oedema, hemorrhage, or infarction) or conflicting native heart and ECMO inflows between the right brachiocephalic and left internal carotid arteries.
There are three potential clinical applications of regional oxygenation saturation (rSO2) during V-A ECMO. First, rSO2 can be used to monitor limb ischemia complications that are related to cannulation,6–8 as Steffen et al.7 reported that limb complications were associated with a nadir rSO2 of <54% and a difference of >35% between the cannulated and noncannulated limbs. Kim et al.8 also reported that no fasciotomy was required for patients who underwent rSO2 monitoring, compared with 13.9% of patients in their control group, and recommended rSO2 monitoring as a useful and reliable method for the early detection of limb ischemia. Second, rScO2 may be used to monitor neurological complications and outcomes, although there are few data regarding rScO2 in the V-A ECMO setting. It is difficult to evaluate the association between neurological complications and rScO2, based on the limited ability to perform computed tomography for patients who are undergoing ECMO. Clair et al.9 are the only researchers so far who have reported that the mean rScO2 was significantly reduced in the right hemisphere of patients with brain injury and death, compared with patients without brain injuries (56 ± 14% vs. 70 ± 8%, p = 0.002). Nevertheless, the positive correlation between neurological outcomes and rScO2 among patients receiving cardiopulmonary resuscitation provides support for the use of rScO2 monitoring of patients during VA-ECMO. Parnia et al.10 performed a multicenter study of patients who experienced in-hospital cardiac arrest and reported that greater than or equal to 60% of cardiopulmonary resuscitation time with a rScO2 value of >50% was useful for predicting a cerebral performance category (CPC) of 1–2 (sensitivity: 77%, specificity: 72%, negative predictive value: 98%). Similarly, Storm et al.11 reported that patients with good neurological outcomes after in-hospital cardiac arrest had significantly higher median rScO2 values (CPC 1–2: 68% vs. CPC 3–5: 58%, p < 0.01). In contrast, our data demonstrated that ROC curves for using rScO2 to discriminate good neurological outcomes of survivor groups showed no statistical difference. The areas under the ROC curves for good neurological outcomes were 0.599 (95% CI: 0.45–0.75, p = 0.153) for right rScO2 and 0.594 (95% CI: 0.45–0.74, p = 0.174) for left rScO2 (data not shown). Nevertheless, as it is difficult to perform computed tomography for patients undergoing ECMO, it is important to pay careful attention to changes in pupil size or light reflex, which may reflect brain complications (edema, hemorrhage, or infarction) among patients with a unilateral decrease in rScO2. Third, rScO2 could be used as a monitoring parameter to predict mortality during V-A ECMO management, as a previous study revealed significant differences in the hemisphere-specific rScO2 values of infants (<3 months old) who survived or died during ECMO (right hemisphere: 69 ± 8% vs. 54 ± 14%, p = 0.0008; left hemisphere: 67 ± 7% vs. 56 ± 16% p = 0.03).9 Similarly, we observed that 28 day mortality was associated with the rScO2 value among adults undergoing ECMO, and the areas under the ROC curves were 0.87 (95% CI: 0.80–0.94, p < 0.001) for right rScO2 and 0.86 (95% CI: 0.80–0.93, p < 0.001) for left rScO2. The optimal cutoff values for right and left rScO2 were 58% (sensitivity: 78.7%, specificity: 83.3%) and 57% (sensitivity: 80.0%, specificity: 70.8%), respectively. Although the small sample size and single-center study design may limit the generalization of these findings, we recommend that clinicians search for hidden problems and consider interventions when patients develop a sudden or marked decrease in their rScO2 value, even if they have otherwise normal hemodynamic parameters.
This study has several limitations. First, the small sample size and single-center design are associated with known risks of bias. Second, the retrospective design may be associated with selection bias, as we only used the cerebral oximeter for relatively serious cases. Third, we did not use computed tomography to confirm whether low rScO2 values were caused by systemic conditions or local brain injuries. Therefore, our findings may not generalize to other centers or patient groups, and a more comprehensive and well-designed prospective study is needed to address these issues.
We suggest that rScO2 may be used as a monitoring parameter for 28 day mortality among patients undergoing V-A ECMO.
This research was supported by the Hallym University Research Fund 2015 (HURF-2015-49). The funding source played no part in the study’s design, data collection and analysis, writing of the manuscript, or the decision to submit the manuscript for publication.
1. Bottiroli M, Decaria D, Maraffi T, et al. Blood lactate normalization following venoarterial ECMO institution for refractory cardiogenic shock. Crit Care 2015.19: 146.
2. Li CL, Wang H, Jia M, Ma N, Meng X, Hou XT. The early dynamic behavior of lactate is linked to mortality
in postcardiotomy patients with extracorporeal membrane oxygenation
support: A retrospective observational study. J Thorac Cardiovasc Surg 2015.149: 1445–1450.
3. Park SJ, Kim SP, Kim JB, et al. Blood lactate level during extracorporeal life support as a surrogate marker for survival. J Thorac Cardiovasc Surg 2014.148: 714–720.
4. Green DW, Kunst G. Cerebral oximetry and its role in adult cardiac, non-cardiac surgery and resuscitation from cardiac arrest. Anaesthesia 2017.72(suppl 1): 48–57.
5. Bickler P, Feiner J, Rollins M, et al. Tissue oximetry and clinical outcomes. Anesth Analg 2017.124: 72–82.
6. Wong JK, Smith TN, Pitcher HT, Hirose H, Cavarocchi NC. Cerebral and lower limb near-infrared spectroscopy in adults on extracorporeal membrane oxygenation
. Artif Organs 2012.36: 659–667.
7. Steffen RJ, Sale S, Anandamurthy B, et al. Using near-infrared spectroscopy to monitor lower extremities in patients on venoarterial extracorporeal membrane oxygenation
. Ann Thorac Surg 2014.98: 1853–1854.
8. Kim DJ, Cho YJ, Park SH, et al. Near-infrared spectroscopy monitoring for early detection of limb ischemia in patients on veno-arterial extracorporeal membrane oxygenation
. ASAIO J 2017.63: 613–617.
9. Clair MP, Rambaud J, Flahault A, et al. Prognostic value of cerebral tissue oxygen saturation during neonatal extracorporeal membrane oxygenation
. PLoS One 2017.12: e0172991.
10. Parnia S, Yang J, Nguyen R, et al. Cerebral oximetry during cardiac arrest: A multicenter study of neurologic outcomes and survival. Crit Care Med 2016.44: 1663–1674.
11. Storm C, Leithner C, Krannich A, et al. Regional cerebral oxygen saturation after cardiac arrest in 60 patients–a prospective outcome study. Resuscitation 2014.85: 1037–1041.
Keywords:Copyright © 2019 by the American Society for Artificial Internal Organs
cerebral oxygenation; extracorporeal membrane oxygenation; mortality