Cerebral Tissue Regional Oxygen Saturation as a Valuable Monitoring Parameter in Pediatric Patients Undergoing Extracorporeal Membrane Oxygenation

Objective: Brain function monitoring technology for extracorporeal membrane oxygenation (ECMO) support has been developing quite slowly. Our objective was to explore the data distribution, variation trend, and variability of cerebral tissue regional oxygen saturation (CrSO2) in pediatric patients undergoing ECMO. Methods: Eight patients who received venoarterial ECMO (V-A ECMO) were included in our study. All of them accepted continuous CrSO2 monitoring by near-infrared spectroscopy (NIRS) within 12 h of ECMO initiation until ECMO wean. Differences in the CrSO2 distribution characteristic, the variation trend of daily CrSO2, and the variability of CrSO2 for the first 5 days following ECMO initiation were compared between survivors and non-survivors according to pediatric intensive care unit (PICU) mortality. Results: The percentage of time of CrSO2 <60% against the whole monitoring time was significantly lower in survivors in both hemispheres {right: 4.34% [interquartile range (IQR) = 0.39–8.55%] vs. 47.45% [IQR = 36.03–64.52%], p = 0.036; left: 0.40% [IQR = 0.01–1.15%] vs. 30.9% [IQR = 26.92–49.62%], p = 0.036}. Survivors had significantly higher CrSO2 on the first 4 days. Root mean of successive squared differences (RMSSD), the variability variable of CrSO2, was significantly lower in survivors (right: 3.29 ± 0.79 vs. 6.16 ± 0.67, p = 0.002; left: 3.56 ± 1.20 vs. 6.04 ± 1.44, p = 0.039). Conclusion: Lower CrSO2, CrSO2 <60% over a longer period of time, and higher fluctuation of CrSO2 are likely associated with PICU mortality in pediatric patients undergoing V-A ECMO. Clinical Trial Registry: URL: http://www.chictr.org.cn/showproj.aspx?proj=46639, trial registry number: ChiCTR1900028021.


INTRODUCTION
Extracorporeal membrane oxygenation (ECMO) is an extracorporeal life support technology used for patients with cardiopulmonary failure as a severe refractory to conventional treatment measure (1)(2)(3). With the development of ECMO, it has been widely applied to the clinical (4-6); however, monitoring technology for ECMO support, particularly the brain function monitoring, has been developing quite slowly.
Near-infrared spectroscopy (NIRS) can provide useful data on cerebral tissue regional oxygen saturation (CrSO 2 ). Because of its characteristics, such as non-invasiveness, continuousness, and being performed in real time and at the bedside (7-10), NIRS has been suggested as one of the regular monitoring technologies during ECMO support by the Extracorporeal Life Support Organization (ELSO) (11). Although NIRS is extensively used in various fields, there have been only a few studies on NIRS in the pediatric patients with ECMO. Consequently, the aim of this study was to explore the data distribution, variation trend, and variability of CrSO 2 in pediatric patients undergoing ECMO so as to provide a relevant reference.

Patients
This prospective observational study was conducted between May 24, 2019, and January 18, 2020 in the 32-bed pediatric intensive care unit (PICU) of the Children's Hospital of Chongqing Medical University, China. All patients aged from 28 days to 18 years needing venoarterial ECMO (V-A ECMO) support were included in this study. Exclusion criteria included the following: with open heart surgery or cardiac catheterization procedure; with ECMO support <24 h; with validated cerebral injury before ECMO, including cerebral hemorrhage and ischemia; with serum total bilirubin >7 mg/dL before ECMO (12); with forehead skin damage not fit for NIRS monitoring.
Assessments of neurologic status were mainly based on the medical history and clinical symptoms and signs before ECMO. Patients suspected of having brain injury were examined with brain computed tomography (CT) or magnetic resonance imaging (MRI). Those patients without suspected brain injury and without abnormal neurologic imaging findings were all considered as without brain injury.
This study has been approved by the Institutional Review Board of Children's Hospital of Chongqing Medical University (2019-237) and has been performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards and has been registered in Chinese Clinical Trial Registry (ChiCTR1900028021). Written informed consent Abbreviations: ECMO, extracorporeal membrane oxygenation; V-A ECMO, venoarterial ECMO; V-V ECMO, venovenous ECMO; PICU, pediatric intensive care unit; NIRS, near-infrared spectroscopy; CrSO 2 , cerebral tissue regional oxygen saturation; SD, standard deviation; CV, coefficient of variation; RMSSD, root mean of successive squared differences; GEE, the generalized estimating equation.
for participation was obtained from children's parents or legal guardians.

Data Collection
Baseline information before ECMO initiation was recorded and included: age, gender, weight, diagnosis, the interval of hospital admission to ECMO initiation, the interval of PICU admission to ECMO initiation, the interval of invasive mechanical ventilation begun to ECMO initiation, the use of vasoactive drugs before ECMO, PaO 2 /FiO 2 ratio, and arterial blood gas parameters (pH, PaCO 2 , PaO 2 , SaO 2 , lactate) before ECMO.
Follow-up variables included ECMO duration, ECMO parameters, treatment measures during ECMO, hemodynamic parameters during ECMO, duration of the invasive mechanical ventilation after ECMO wean, length of PICU and hospital stays, and PICU mortality.

ECMO Management
All patients receiving ECMO support were managed by our ECMO team. We chose the mode of ECMO support in pediatric respiratory failure according to the patient's hemodynamic status and weight. We have to choose V-A ECMO in infants and low-weight toddlers with respiratory failure purely for a lack of double-lumen venovenous (V-V) cannulas in China. Venoarterial ECMO was surgically implanted using heparincoated cannulation (Bio-Medicus, Medtronic Inc., Minneapolis, MN, USA) via the right internal jugular vein and the right common carotid artery. The centrifugal pump (Deltastream DP3, Medos, Stolberg, Germany) and membrane oxygenator (Hilite 800 LT, Hilite 2400 LT, or Hilite 7000 LT, Medos, Stolberg, Germany) were used. ECMO-related parameters and hemodynamic parameters were recorded every 2 h during ECMO support.

Cerebral Tissue Regional Oxygen Saturation Monitoring
Monitoring of CrSO 2 began within 12 h of ECMO initiation and continued until the weaning of ECMO using NIRS (FORE-SIGHT P/N01-06-2030C, CAS Medical Systems Inc., Branford, CT, USA). Cerebral tissue regional oxygen saturation data were gathered every 2 s and recorded on the device's memory. Our ECMO team did not adjust therapies following CrSO 2 data. The daily median CrSO 2 values were calculated for the first 5 days following ECMO initiation in both hemispheres and were then used to analyze the variation trend of CrSO 2 . At the same time, the median CrSO 2 values over 4-h time intervals were calculated for the first 5 days and plotted as a time series. The variability of CrSO 2 was evaluated using the standard deviation (SD), coefficient of variation (CV), and root mean of successive squared differences (RMSSD).
where x i is the median CrSO 2 at interval i (13). The differences of different CrSO 2 variables were compared between survivors and non-survivors.

Outcome
The primary outcome was PICU mortality. All patients weaned off ECMO successfully had accepted brain CT or MRI as soon as the patients' condition relatively stabilized.

Statistical Analysis
Statistical analysis was performed with SPSS version 26 (IBM Corp., Armonk, NY, USA). Continuous variables were confirmed for normal distribution using the Shapiro-Wilk test and are presented either as means ± SD for data with a normal distribution or as median (IQR) for nonnormal data. The differences between groups were tested with Student t-test, Satterthwaite t-test, or Mann-Whitney U-test, as appropriate. Categorical variables are presented as counts (percentages). Repeated measurement data were analyzed using the generalized estimating equation. p < 0.05 was considered as statistically significant.

Cerebral Tissue Regional Oxygen Saturation Variables
The duration of CrSO 2 monitoring was 79.0 ± 17.0% of the time on the first 5 days of ECMO. The distribution of CrSO 2 data for the first 5 days in both hemispheres is shown in Figure 1.  Figure 2 show CrSO 2 variation trends for the first 5 days in both hemispheres. Because of the risk of developing pressure sores, we did not gather CrSO 2 signals in the fifth day for patient 1. While the duration of ECMO support was only 41 h for patient 8; thus, the patient did not contribute data to the third, fourth, and fifth day. Survivors had significantly higher CrSO 2 for each of the first 4 days in both hemispheres (all p < 0.05), but there were no significant differences between survivors and nonsurvivors on the fifth day in both hemispheres (right: Wald χ 2 = 3.402, p = 0.065, left: Wald χ 2 = 1.216, p = 0.270).
The indicators of CrSO 2 variability between survivors and non-survivors for the first 5 days following ECMO initiation are shown in

DISCUSSION
Monitoring technology has not been developed along with ECMO-related technologies, and monitoring of the central nervous system is particularly challenging. The head ultrasound was used in patients undergoing ECMO with several limitations (16,17), for example, this approach could not capture the neurologic imaging in pediatric patients with closed cranial fontanels and adult patients (8). Besides, it cannot be used for continuous monitoring. As for electroencephalography, amplitude-integrated electroencephalography, and somatosensory-evoked potentials, a few studies have tested their roles in ECMO support (8,18). Although head CT and MRI can clearly detect the cerebral hemorrhage or ischemia, the hospital transfer may have potential risks (8). Different from all these monitoring methods, cerebral NIRS can provide a continuous value of CrSO 2 . Because of its non-invasiveness, continuousness, and being performed in real time and at the bedside (7-10), ELSO suggests regular application of NIRS to patients undergoing ECMO support (11). Nonetheless, there was not a certain method to use NIRS. In our study, we explored the CrSO 2 distribution characteristic, the variation trend of daily CrSO 2 , and the variability of CrSO 2 for the first 5 days following ECMO initiation.
In the present study, survivors had higher CrSO 2 for the first 4 days following ECMO initiation in both hemispheres, and the percentage of time of CrSO 2 <60% against the whole monitoring time was lower in survivors. Previous studies also revealed that survivors had higher CrSO 2 than non-survivors (19,20). And the occurrence of cerebral desaturation, which is defined as CrSO 2 <60% for >5% of the entire monitoring period, was independently associated with hospital mortality (21). Wong et al. (22) adjusted therapy measures that could affect the cerebral oxygen balance to improve CrSO 2 values. They also found that unilateral or bilateral CrSO 2 decreasing persistently was associated with worse outcome. We have similar findings and consider that lower CrSO 2 and the longer time of CrSO 2 <60% reduce the possibility of survival. However, there was no significant difference for the median CrSO 2 of the fifth day among survivors and non-survivors. We guess volume overload may play a role in this phenomenon, because all deceased patients in our study accepted continuous renal replace therapy in the later days of ECMO support. Studies suggested cerebral autoregulation impairment during ECMO was associated with worse outcome (20,23). Excessive blood would stream into brain circulation without the regulation of cerebral autoregulation, so that the CrSO 2 signal gets higher.
Besides, for the indicators of CrSO 2 variability, bilateral RMSSD and left CV were significantly higher in non-survivors; meanwhile, bilateral SD and right CV also showed the trend to be higher in non-survivors. Therefore, we speculate that greater fluctuation of CrSO 2 is associated with a worse prognosis. Absolute CrSO 2 values can be different between different types of NIRS devices, because of different algorithms (24). Different from using absolute CrSO 2 value directly, indicators of CrSO 2 variability can reduce the error caused by different types of NIRS devices. Using RMSSD to describe the variability of CrSO 2 had been studied in perioperative period of cardiac surgery. Spaeder et al. (13) found that lower RMSSD was associated with poor neurodevelopmental outcomes in neonatal survivors of congenital heart disease. Meanwhile, Flechet et al. (25) found that pediatric patients with lower RMSSD had larger probability of developing acute kidney injury after cardiac surgery. This is the first study to explore the CrSO 2 variability using RMSSD in patients with ECMO support. However, different from those findings that came from the perioperative period of cardiac surgery, survivors had a lower RMSSD in our study. We consider that there are several potential factors that may cause this difference. First, patients with open heart surgery were excluded in our study, whereas those two previous studies had focused on the perioperative period of cardiac surgery. Second, conclusion of those two previous studies came from CrSO 2 data gathered from postoperative period. Cerebral blood flow is pulsatile for most patients in this period, whereas cerebral blood flow in V-A ECMO is non-pulsatile or partial pulsatile, which is associated with endothelial dysfunction, decreased microvascular perfusion, and increased vascular resistance (26). Third, the total monitoring duration is longer in our study. Meanwhile, those two previous studies averaged CrSO 2 over 1-min intervals to calculate RMSSD. But we calculated RMSSD using the median CrSO 2 over 4-h intervals, because missing data cannot be avoided in such long a time in clinical. We have checked CrSO 2 signal every 4-6 h, so that we have enough data in every 4-h intervals. Further studies are needed to confirm our findings and explore underlying mechanisms.
There are some limitations to our study. First, our findings came from a small sample size and single-center study, and most patients in our study had a similar diagnosis of human adenovirus pneumonia. Those make it cautious to interpret the results of our study. Second, only patients receiving V-A ECMO were included in our study; thus, conclusions obtained from the current study may not be extrapolated to those with V-V ECMO. However, Clair et al. (19) reported the mean CrSO 2 for the whole time of ECMO support, the duration of CrSO 2 below the threshold of 20% from baseline, and the duration of CrSO 2 <50% were not different in patients with V-V ECMO compared to V-A ECMO. Third, we only compared CrSO 2 values between survivors and non-survivors, as the small sample size is a limitation in adjusting for suspected confounding factors. The relationships between CrSO 2 values and hemodynamic parameters were not discussed. At the same time, we did not explore the correlation between CrSO 2 and neurologic complications. Studies found that lower CrSO 2 and more frequently occurrence of cerebral desaturation in patients with ECMO were more likely related to brain injury (19,27,28). However, different studies had different definitions of cerebral desaturation and brain injury. Tian et al. (29) assessed the dynamic relationship between the MAP and CrSO 2 fluctuations and found that patients with significant in-phase MAP to CrSO 2 coherence always had abnormal neurologic imaging findings. Finally, there are limitations in CrSO 2 monitoring. CrSO 2 mainly represents the regional oxygen saturation of the forehead. Thus, it may be hard to evaluate the deeper part or other region of the brain. Meanwhile, CrSO 2 can be affected by serum bilirubin, skin surface temperature, or scalp blood flow (12,21,28). CONCLUSION Cerebral tissue regional oxygen saturation can be used as a regular monitoring parameter for pediatric patients undergoing ECMO. Lower CrSO 2 , CrSO 2 <60% over a longer period of time, and higher fluctuation of CrSO 2 are likely associated with PICU mortality in pediatric patients undergoing V-A ECMO. Future multicenter and large-scale studies are needed to mitigate confounding factors and verify our findings and explore the correlation between neurologic complications and CrSO 2 variables in our study.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Institutional Review Board of Children's Hospital of Chongqing Medical University. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

AUTHOR CONTRIBUTIONS
SC is the principal author and contributed to data collection, data analysis and interpretation, and drafted the manuscript. SC and FX contributed to the study conception and design. FF, WL, CL, and FX contributed to data analysis and interpretation and revised the manuscript critically for important intellectual content. All authors contributed to the article and approved the submitted version.