Prevalence and Neurological Outcomes of Comatose Patients with Extracorporeal Membrane Oxygenation

Abstract

Objective:

To investigate prevalence, risk factors, and in-hospital outcomes of comatose extracorporeal membrane oxygenation (ECMO) patients.

Design:

Retrospective observational.

Setting:

Tertiary academic hospital.

Participants:

Adults received venoarterial (VA) or venovenous (VV) ECMO support between 11/2017–04/2022.

Interventions:

None.

Measurements and Main Results:

“24-hour off sedation” was defined as no sedative infusion (except dexmedetomidine) or paralytics administration over a continuous 24-hour period while on ECMO. “Off-sedation coma” (coma_off) was defined as GCS ≤8 after achieving 24-hour off sedation. “On-sedation coma” (coma_on) was defined as GCS≤8 during the entire ECMO course without off-sedation for 24 hours. Neurological outcomes were assessed at discharge using the modified Rankin scale (good:0–3, poor:4–6).

230 patients were included (VA-ECMO 143, male 65%). “24-hour off sedation” was achieved in 32.2% VA-ECMO and 26.4% VV-ECMO patients. Among all patients off sedation for 24 hours (n=69), 56.5% VA-ECMO and 52.2% VV-ECMO patients experienced coma_off. Among those unable to be sedation-free for 24 hours (n=161), 50.5% VA-ECMO and 17.2% VV-ECMO had coma_on. Coma_off associated with poor outcomes (p<0.05) in VA-ECMO and VV-ECMO groups while coma_on only impacted the VA-ECMO group outcomes. In a multivariable analysis, requirement of renal replacement therapy (RRT) was an independent risk factor for coma_off after adjusting for ECMO configuration, after adjusting for ECMO configuration, acute brain injury (ABI), pre-ECMO PaO₂, PaCO₂, PH and bicarbonate level (worst value within 24 hours before cannulation).

Conclusions:

Coma_off was common and associated with poor outcomes at discharge. Requirement of RRT was an independent risk factor.

INTRODUCTION

Since the first successful application of extracorporeal membrane oxygenation (ECMO) in patients in the 1970s, ECMO has become an increasingly utilized lifesaving technique for patients with refractory cardiopulmonary failure ^1,2. It serves as a bridge to recovery, other mechanical circulatory support device, or heart/lung transplant ^3,4. It has been reported that close to 50% of the adult patients with ECMO survive to hospital discharge ⁵. However, acute brain injury (ABI) diagnosed during ECMO support is a major contributing factor to poor functional outcomes and mortality ⁶. ABI includes intracranial hemorrhage (ICH), ischemic stroke, seizure, hypoxic-ischemic brain injury (HIBI), brain death, and cerebral edema ⁷. ABI has been reported in 7.7% of the patients with venoarterial (VA)-ECMO (non-extracorporeal cardiopulmonary resuscitation) (ECPR) and 7.1% of the patients with venovenous (VV)-ECMO ^8,9. In the most recent Extracorporeal Life Support Organization Registry analysis, ABI occurred in 16.5% ECPR patients ⁸. The reported incidence of ABI is higher in autopsy studies, likely due to underdiagnosis in patients with ECMO with underutilization of neuroimaging studies ^10,11.

Persistent coma in patients with ECMO may be a unique manifestation of clinically significant ABI, the effects of metabolic derangement without identifiable structural brain injuries, or simply a consequence of sedation. As ECMO patients with commonly have multi-organ failure and receive prolonged sedation, differentiating whether coma is from significant ABI or sedation can be challenging. Notably, in one study, 12.6% (11 of 87) of the ECMO patients had unexplained coma without relevant imaging findings, which accounted for 26% of all ABI patients ¹². In an autopsy study of 4 patients treated with ECMO who had persistent coma, in 2 of the 4 cases, the comatose status was not explained by the neuropathology findings ¹³. Despite these observations and clinical experience, the prevalence of coma and its clinical implication in ECMO patients are unknown. There are no reports on the in-hospital outcomes of comatose ECMO patients. And yet, the presumption of poor neurological outcomes due to persistent coma in ECMO patients is frequently used to guide end-of-life discussions in patients with ECMO ^14,15. Here, we aimed to better define the neurological outcomes of ECMO-associated coma. Our study investigated the prevalence, risk factors, and in-hospital outcomes of comatose ECMO patients. We hypothesize that persistent coma when off sedation is an independent predictor of poor hospital discharge outcomes.

METHODS

Study Design

A retrospective observational cohort study was conducted on patients with VA- and VV-ECMO at a tertiary medical center. All adults (age ≥ 18) who received ECMO support between November 2017 and April 2022 were included. All patients were admitted to the cardiovascular surgical or cardiac intensive care unit and were followed by the neurocritical care team adhering to our standardized neuromonitoring protocol (supplemental material) from day 1 post ECMO cannulation until hospital discharge or death ⁶. This study adhered to the Declaration of Helsinki principles, local legislation, and institutional requirements. It was approved by the Johns Hopkins Medicine Institutional Review Board. Informed consent was obtained from all participants via their legally authorized representatives (IRB00216321: An Observational Registry of Patients who Received Extracorporeal Membrane Oxygenation at Johns Hopkins Hospital, approved on 12/28/2020).

Definitions

“24-hour off sedation” was defined as no sedative infusion (except dexmedetomidine) or paralytics administration over a continuous 24-hour period while on ECMO. Pro Re Nata (PRN) boluses with sedatives and/or analgesics were allowed. Glasgow Coma Scale (GCS) was assessed every 1–4 hours.

“Off-sedation coma” or “coma_off” was defined as GCS remained ≤8 during the first 24 hour period off sedation. If a patient had one GCS recording ≥9 during that 24-hour period, this patient was considered as “off-sedation non-coma” or “non-coma_off”. “On-sedation coma” or “comaon” was defined as GCS≤8 during the entire ECMO course without being off sedation for 24 hours. If a patient had one GCS recording ≥9 during the same period, this patient was considered as “on-sedation non-coma” or “non-coma_on”.

We defined ABI as any newly diagnosed acute neurological insult during ECMO support, including ischemic stroke, ICH, subdural hematoma (SDH), subarachnoid hemorrhage (SAH), seizure, HIBI, brain death, cerebral edema, and central nervous system (CNS) infection.

Clinical Protocol and Outcomes

Neurological exams were performed at least daily by the primary team and/or the neurocritical care consult team, and the exams were conducted when patients were on a minimal amount of sedation or completely off sedation. In line with our institution’s sedation protocol, efforts to wean patients off sedating medications occurred daily, and sedation holidays were routinely administered when deemed safe ⁶.

All cardiovascular intensive care unit (CVSICU) nurses received comprehensive training from neurocritical care nurses on coma exams and sedation management, emphasizing minimal sedation to facilitate neurological assessment ^6,16. All ECMO patients are closely monitored by neurocritical care fellows/attendings on day 1 post cannulation. The CVSICU nurses were directed to consult the neurocritical care notes for accurate documentation.

For multimodal neuro-monitoring, nearly all patients underwent at least one transcranial Doppler study to detect microemboli, measure mean flow velocity, and assess pulsatility. Other studies, including computer tomography (CT), electroencephalogram (EEG) and somatosensory evoked potentials (SSEP), were performed as clinically indicated. Typically, no additional neuromonitoring was pursued for awake patients without focal neurological deficits. If the patients remained comatose, EEG and CT scans were usually obtained.

Patients supported with either VA-ECMO or VV-ECMO were divided into 4 subgroups, based on off-sedation status and GCS scores (Figure 1): off-sedation coma (coma_off), off-sedation non-coma (non-coma_off), on-sedation coma (coma_on), and on-sedation non-coma (non-coma_on). For those in coma_off group, we continued to track GCS until hospital discharge or death and recorded the best GCS.

Figure 1:

Flowchart of the study design.

The primary outcome was the patient’s neurological functional status at discharge: categorized as good for a modified Rankin scale (mRS) ≤3 and poor if mRS ≥4. The mRS was determined by a physical therapist. We compared the outcomes between comatose patients and non-comatose patients, and explored potential associations of relevant variables (both pre-ECMO and post-ECMO) with the coma status.

Withdrawal of life-sustaining therapy (WLST) was a collective decision involving cardiothoracic surgeons, CVSICU intensivists, palliative care professionals, and the patient’s family. If significant neurological injuries were present, neuro-intensivists were also involved. Considerations encompassed the patient’s quality of life pre-ECMO, comorbidities, complications such as ABIs during hospitalization, age, personal wishes, family beliefs, and the anticipated neurological outcomes.

Statistical Analysis

We did not embark upon a prior sample size and power calculations but rather intended to include all ECMO patients cared for in our institution during the study period. Demographic and clinical data were reported as the number of counts with percentage or the median with interquartile range (IQR). Those variables between patients with or without coma were compared using Wilcoxon rank-sum test for continuous variables and Fisher’s exact test for binary or categorical variables.

We conducted Kaplan-Meier (KM) survival analysis using the log-rank test to compare differences between data subsets. Neurological outcomes at discharge were compared between the coma_off and non-coma_off patients. ECMO cannulation date was set as time zero, and patients were followed until hospital discharge or death. Additionally, we performed Cox proportional hazards regression analysis on the same dataset, confirming the proportional hazards assumption with Schoenfeld residuals, which revealed no significant violations.

Univariate logistic regression tests were performed to analyze the dichotomized outcomes (coma_off and non-coma_off). We selected 22 pre-defined covariates (Table 2). We then built a multivariable logistic regression statistical model. Selected clinically relevant covariates (Table 3) were included to identify risk factors for sustained coma when off sedation. A two-way interaction analysis was conducted on the selected covariates. Two-tailed P value < 0.05 was considered statistically significant. Odds ratios and 95% confidence intervals were calculated. STATA version 18.0 (STATA Corp, College Station, Texas, USA) was used for statistical analysis. Prism 9.5.1 (GraphPad Software, Boston, MA, USA) was used to generate tables and graphs.

Table 2:

Univariable logistic regression analysis of off-sedation coma (coma_off) in patients with VA and VV ECMO.

Univariable Analysis: Comaoff
Variables	OR	95% CI	p value
Pre-ECMO worst PCO2	0.99	0.97–1.01	0.421
Pre-ECMO worst PO2	1.00	1.00–1.00	0.623
Pre-ECMO worst bicarb	0.93	0.84–1.02	0.11
Pre-ECMO worst PH	0.03	0.00–2.04	0.102
ECMO type	1.19	0.44–3.25	0.732
Bacteremia	1.17	0.40–3.40	0.772
BMI	1.11	1.02–1.21	0.011
Age	1.02	0.98–1.06	0.262
Gender	0.73	0.27–1.98	0.536
Race	1.27	0.83–1.94	0.276
RRT	4.39	1.43–13.50	0.010
Arterial line MAP	0.93	0.86–1.02	0.119
Covid 19	2.09	0.58–7.60	0.261
HTN	1.53	0.53–4.43	0.429
CHF	0.32	0.11–0.93	0.036
DM	1.17	0.40–3.40	0.772
HLD	2.17	0.83–5.73	0.115
CKD	1.52	0.40–5.78	0.535
A-fib	0.55	0.18–1.70	0.302
ABI	2.1	0.78–5.62	0.140
AST	1.00	1.00–1.00	0.045
ALT	1.01	1.00–1.01	0.042

RRT renal replacement therapy, HTN hypertension, CHF chronic heart failure, DM diabetes, HLD hyperlipidemia, CKD chronic kidney disease, A-fib atrial fibrillation, ABI acute brain injury, MAP mean arterial pressure.

Table 3:

Multivariable logistic regression analysis of off-sedation coma (coma_off) in patients with both VA and VV ECMO.

Multivariable Analysis: Comaoff
Variables	aOR	95% CI	p value
ECMO type	0.57	0.11–3.06	0.52
ABI	2.23	0.75–7.01	0.15
Pre-ECMO worst PH	0.15	0.00–1443	0.68
Pre-ECMO worst bicarb	0.90	0.72–1.13	0.36
Pre-ECMO worst PCO2	0.99	0.92–1.06	0.73
Pre-ECMO worst PO2	1.00	0.99–1.00	0.75
RRT	4.37	1.21–15.81	0.024

RRT renal replacement therapy, ABI acute brain injury.

RESULTS

Baseline Characteristics

A total of 230 adult patients were included in the cohort. Of those, 143 (62%) were on VA-ECMO support and 87 (38%) were on VV-ECMO support. The baseline demographics and characteristics of the cohort are shown in Table 1. Among the patients with VA-ECMO, 64.3% were males with the median age of 58 years (IQR=45–68). In the VV-ECMO group, 66.7% were males with the median age of 48 years (IQR=39–56). The median duration of ECMO support was 5.4 days in the VA-ECMO group and 21.1 days in the VV-ECMO group. Please refer to Supplementary Figure 1 for more details.

Table 1.

Baseline characteristics of patients undergo VA or VV ECMO.

Characteristics	VA-ECMO patients n=143	VV-ECMO patients n=87
Demographics
Age (years), median (IQR)	58 (45–68)	48 (39–56)
Male, n (%)	92 (64.3%)	58 (66.7%)
BMI, kg/m2 (IQR)	29.4 (25.0–35.6)	33.2 (28.2–37.3)
Race (W/B/H/A/Other)	82/45/3/5/8	32/29/20/3/3
Comorbidities
Hypertension	103	35
Diabetes	47	14
Hyperlipidemia	78	29
Chronic renal disease	23	3
Atrial fibrillation	39	0
Congestive heart failure	48	0
Prior ischemic stroke	14	4
Prior hemorrhagic stroke	1	0
Hospital stay (days), median (IQR)	20 (8–47)	49 (24–77)
ECMO length (days), median (IQR)	5.4 (2.9–8.8)	21.1 (10.2–42.3)
VA-ECMO Indications (top 3)
Cardiogenic shock	74
Post cardiotomy shock	43
Cardiac arrhythmia	23
VV-ECMO indications (top 2)
COVID-19 pneumonia		54
Bacterial pneumonia		8

ECMO extracorporeal cardiopulmonary oxygenation; IQR interquartile ratio; W white; B black; H Hispanic; A Asian; BMI body mass index

VA-ECMO

Coma_off

A continuous 24-hour off-sedation period was achieved in 46 patients (32.2%) with a median off-sedation duration of 2.5 days (IQR=2.0–5.0). Among those patients, 26 (56.5%) patients remained GCS≤8 (coma_off) while 20 (43%) patients recovered to at least one GCS score ≥9 (non-coma_off). The coma_off group had significantly fewer good neurological outcomes compared with the non-coma_off group (n=0, 0%, vs. n=6, 30%, p=0.004) (Figure 2A). The KM survival curve also demonstrated that the coma_off group had significantly lower probability of good neurological outcomes (p=0.0001) (Figure 2B). Cox proportional hazards regression analysis also showed a similar finding that non-coma_off was significantly associated with a decreased hazard of poor outcome at hospital discharge (p=0.0006). Among the 26 patients with coma_off, 2 patients (7.7%) had an improvement in GCS score to ≥9 during the hospital course. However, both had poor neurological outcomes at discharge. The one with mRS of 5 at discharge was accepted into a nursing home. The other one with mRS of 6 underwent WLST.

Figure 2.

Coma_off, hospital discharge outcome, and acute brain injury (ABI) in the patients with VA-ECMO. A, the relationship between coma_off and the outcomes at discharge. B, Kaplan Meier survival analysis comparing the probability of good outcome between the coma_off and non-comao_ff subgroups at hospital discharge. Each solid dot represents a censored subject. C, the relationship between ABI and coma_off. D. Different types of ABI in the coma_off patients.

Among the 46 patients who achieved 24-hour off sedation, 20 (43.5%) had at least one type of ABI. There was a non-statistically significant trend towards higher prevalence of ABI in coma_off patients vs. non-coma_off patients (n=14, 53.8%, vs. n=6, 30.0%, p=0.14) (Figure 2C). Among the 14 coma_off patients with ABI, ischemic stroke and HIBI were the most common types (Figure 2D).

Coma_on

In our cohort, 97 patients never achieved 24-hour off sedation while on ECMO. 48 (49.5%) had at least one GCS score ≥9 (non-coma_on) and 49 (50.5%) remained GCS ≤8 (coma_on) during the ECMO course. The coma_on group had significantly fewer good neurological outcomes compared with the non-coma_on group (n=3, 6.1%, vs. n=21, 43.8%, p<0.0001) (Supplementary Figure 2A). The prevalence of ABI in the coma_on and the non-coma_on groups can be found in Supplementary Materials and Supplementary Figure 2B.

VV-ECMO

Coma_off

A continuous 24-hour off sedation period was achieved in 23 (26.4%) patients undergoing VV ECMO with the median off-sedation duration of 3.0 days (IQR=2.0–9.0). Among those patients, 12 (52.2%) remained GCS≤8 (coma_off) while 11 (47.8%) had recovered to at least one GCS score ≥9 (non-coma_off). The coma_off group had significantly fewer good neurological outcomes compared with the non-coma_off group (n=1, 8.3%, vs. n=7, 63.6%, p=0.009) (Figure 3A). The KM survival curve also demonstrated the coma_off group had significantly lower probability of good neurological outcomes (p=0.0003) (Figure 3B). Cox proportional hazards regression analysis also showed a similar finding that non-coma_off was significantly associated with a decreased hazard of poor outcome at hospital discharge (p=0.0033). Among the 12 coma_off patients, 2 (16.7%) had an improvement in GCS score to ≥9. However, both eventually went through WLST.

Figure 3.

Coma_off, hospital discharge outcome, and acute brain injury (ABI) in the patients with VV-ECMO. A, the relationship between coma_off and the outcomes at discharge. B, Kaplan Meier survival analysis comparing the probability of good outcome between the coma_off and non-comao_ff subgroups at hospital discharge. Each solid dot represents a censored subject. C, the relationship between ABI and coma_off. D. Different types of ABI in the coma_off patients.

Among the 23 patients who achieved 24-hour off sedation, 9 (39.1%) had at least one type of ABI. The prevalence of ABI was similar between the coma_off patients vs. the non-coma_off patients (n=5, 41.8, vs. n=4, 36.4%, p>0.99) (Figure 3C). Among the 5 coma_off patients with ABI, ICH and SAH were the most common types (Figure 3D).

Coma_on

Of the 64 patients who never achieved 24-hour off sedation while on ECMO, 53 (82.8%) had at least one GCS score ≥9 (non-comaon) and 11 (17.2%) remained GCS ≤8 (coma_on) during the ECMO course. The coma_on group had fewer good neurological outcomes compared with the non-coma_on group (n=3, 27.3%, vs. n=22, 41.5%, p=0.50) (Supplementary Figure 2C). The prevalence of ABI in the coma_on and the non-coma_on groups can be found in Supplementary Materials and Supplementary Figure 2D.

Risk Factors for Coma_off

In the multivariable analysis statistical model, requirement of renal replacement therapy (RRT) was a significant risk factor for coma_off after adjusting for ECMO configuration, acute brain injury (ABI), pre-ECMO PaO₂, PaCO₂, PH and bicarbonate value (worst value within 24 hours before cannulation) (Table 3). A two-way interaction analysis was conducted on all 7 selected covariates and there was no significant interaction effect between any 2 covariates.

Sedation Practice Before and After the Covid Pandemic

Finally, we explored the possibility that the COVID-19 pandemic may have changed our sedation practice in the ICU since our cohort comprised of patients from 2017 to 2022. March 2020 was considered as the beginning of the pandemic. No significant difference was found between the percentage of patients who achieved 24-hour off-sedation in VA-ECMO group (p=0.72) or VV-ECMO group (p=0.55) before and after the COVID-19 pandemic (Supplemental Figure 3A-B).

Mortality and Withdrawal of Life-Sustaining Treatments (WLST)

The overall mortality for the patients with VA-ECMO was 65.7% (94/143) and 48.3% (42/87) in the patients with VV-ECMO. In the individual subgroups, the mortality was 88.5% (VA-ECMO coma_off), 50.0% (VA-ECMO non-coma_off), 83.7% (VA-ECMO coma_on), 41.7% (VA-ECMO non-coma_on), 75.0% (VV-ECMO coma_off), 18.2% (VV-ECMO non-coma_off), 63.6% (VV-ECMO coma_on) and 45.3% (VV-ECMO non-coma_on) (Supplementary Figure 3C-D).

Since persistent coma was often quoted to support the decision for WLST, we specifically examined WLST vs natural death in both VA- and VV- ECMO groups. We found no significant difference between these two groups (p = 0.5836) (Supplementary Figure 3E). Similarly, when comparing WLST decisions between the VA-ECMO coma_off and non-coma_off subgroups, and between the VV-ECMO coma_off and non-coma_off subgroups, no significant difference was observed (p = 0.5343 and p>0.9999, respectively, as shown in Supplementary Figure 3F and 3G). Overall, our data indicate that persistent coma while off sedation did not significantly influence the WLST decisions.

Discussion

Our study is the first comprehensive analysis of prevalence, risk factors, and neurological outcomes of comatose ECMO patients. We found that coma_off was common in ECMO patients. Requirement of RRT during ECMO support was independent risk factor for coma_off. Coma_off was strongly associated with unfavorable neurological outcomes at hospital discharge. There are multiple unique strengths of our study: 1) By implementing the institutional sedation cessation protocol, we minimized the confounding effect of sedation on the coma exams whenever feasible ⁶; 2) With the sedation cessation protocol in place, we were able to compare the patients on-sedation vs. those off-sedation; and 3) A standardized neuromonitoring protocol allowed for an early detection of ABIs in ECMO patients⁶.

Coma was commonly diagnosed when standardized sedation cessation and neuromonitoring protocols were implemented. In our cohort, the coma prevalence was: 56.6% in the VA-ECMO off-sedation group, 50.5% in the VA-ECMO on-sedation group, 52.2% in the VV-ECMO off-sedation group, and 17.2% in the VV-ECMO on-sedation group. The prevalence of coma has been reported in other ICU patient populations. Among out-of-hospital cardiac arrest survivors (unclear if off sedation), 56% remained in coma for > 24 hours after achieving return of spontaneous circulation, compared to 30% of in-hospital cardiac arrest survivors ¹⁷. Approximately 82% (however, <1% off sedation) of the COVID-19 ICU patients remained comatose for a median of 10 days ¹⁸. Coma was also found in 16% of the patients (all off sedation) with severe sepsis ¹⁹. But none of those studies implemented a standardized sedation cessation protocol as the one used in our cohort. It is worth mentioning that awake ECMO have been attempted with both VA- and VV-ECMO. The definition of awake varies in different studies. “Awake ECMO” was defined as invasive mechanical ventilation used during ≤ 50% of the VA ECMO run ²⁰. It was defined as cannulation done while the patients were awake and spontaneously breathing without endotracheal tube ²¹. Awake ECMO is further combined with physical therapy and in a systemic review, ambulation was surprisingly achieved in 43% of the patients with awake ECMO ²².

Coma was associated with higher mortality and poor neurological outcomes. In our cohort, more patients died in the VA-ECMO coma_off group than the non-coma_off group (88.5% vs. 50.0%). Similar trend was also found in the VV-ECMO off-sedation patients (75.0% vs. 18.2%). The same pattern was observed in the neurological outcomes of those patients. It is surprising that no patient in the VA-ECMO coma_off group and only 1 patient in the VV-ECMO coma_off group had good outcomes upon discharge. Only 4 patients with comaoff regained some consciousness (GCS≥9) later in the hospital course, but they all had poor neurological outcomes at discharge, indicating that coma_off may represent a surrogate marker for poor outcomes. The association of coma and mortality was also reported in other patient populations, such as post cardiac arrest (unclear if off sedation), ARDS related to COVID-19 (not off sedation), and severe sepsis (off sedation) ^17,19,23. Because of this significant association, being comatose while off sedation is commonly used to facilitate the decision to withdraw care in the real-world practice although it’s difficult to determine the “sufficient” off-sedation time. Also, it is important to remember that there is a high risk of bias from the self-fulfilling prophecy ¹⁵. In our cohort, WLST contributed to 86.2% of death in the VA-ECMO group and 90.1% in the entire VV-ECMO group.

Coma in the ICU setting is frequently multifactorial in nature. Common etiologies include ABI, sedation/toxins, metabolic/electrolyte/endocrine derangement, nutrition, temperature, and infection ²⁴. We included clinically relevant variables for coma_off in our regression model and identified one independent risk factors: RRT. In our patient cohort, majority of our patients (58.3%) received RRT. Those requiring RRT were sicker, had more metabolic derangements and their pharmacokinetics was also very different. ABI was discovered in 31.7% of the ECMO patients in our study, which is likely due to our standardized neuromonitoring protocol ¹⁶. However, it was not significantly associated with coma_off. in our multivariable regression analysis. The average levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) during the ECMO course were elevated. But their impact on the coma status was minimal if any with an unadjusted OR of only 1.00 and 1.01, respectively. Sedation frequently confounds coma exams ¹⁵, but proper sedation is often required to reduce ventilator dyssynchrony and to facilitate therapeutic interventions in the ICU ²⁵. We allowed dexmedetomidine infusion during the 24-hour off sedation period because of its capacity to maintain consciousness or awareness in the ICU patients ²⁶. We implemented a sedation cessation protocol to all ECMO patients and minimized sedation when deemed safe. It is worth emphasizing that the median off-sedation duration was 2.5 days (IQR: 2.0–5.0) for the VA-ECMO off-sedation group, and 3.0 days (IQR: 2.0–9.0) for the VV-ECMO off-sedation group, which may be “sufficient” time to diagnose coma_off. Although our study was not able to adjust for all clinically relevant variables, this is by far the most controlled coma study in the ECMO literature.

It is interesting to note that a higher percentage of patients were comatose while off sedation compared to those on sedation. Patients who were awake while on ECMO were more likely to express discomforts or their discomforts were more likely to be noticed by their health care providers. As a result, those patients were likely to receive more sedatives, which reduced the number of non-coma_off patients and increased the number of non-coma_on patients. Conversely, the comatose patients were less likely to display discomforts, leading to the administration of fewer sedatives. Consequently, the percentage of total patients with coma_off (55.1%) was higher than that of patients with coma_on (37.3%).

Compared with the adult ELSO database, our cohorts had a higher mortality rate. In our center, we consider ECMO as a true rescue therapy and therefore we tend to delay cannulation until ECMO support is absolutely necessary²⁷. As shown in our experience with COVID-19, our center delayed initiation of VV-ECMO support, which correlated with a higher mortality rate compared to other centers ²⁸. In addition, 29.9% patients in our cohort were in postcardiotomy shock, which usually carries a higher mortality rate than the general patients with VA-ECMO ²⁹.

Limitations

Our study also has several limitations. 1) It is a retrospective single-center study. The data may not be generalizable to other institutions and populations. 2) Although we have a total of 230 patients treated with ECMO, the sample size in some subgroups is in the lower teens or even in single digits. Therefore, the study may not have enough power to detect contributions of various risk factors to the comatose status. In addition, due to missing data, many clinically important variables were not included in the analyses, such as sequential organ failure assessment (SOFA) and acute physiology and chronic health evaluation (APACHE) II scores. However, we have pre-ECMO worst PCO2, pre-ECMO worst PO2, pre-ECMO worst bicarbonate level, pre-ECMO worst PH, requirement of RRT, and presence of ABI in the model to adjust the severity of illness. In addition to those variables, in the univariate analysis, we also included liver enzymes, presence of bacteremia, blood pressure etc. 3) Only a fraction of our patients achieved the coma_off state, and the coma_on state presented significant confounders for coma assessment. These may further limit the generalizability of our study across ECMO patients. But this study presents a viable method to further define the impact of coma on ECMO. 4) We did not have enough data to explore the association between coma_off and WLST and to investigate the possible self-fulfilling prophecy. Further studies are needed to answer this important question. 5) In an ideal scenario, the cumulative dosage of sedatives prior to the “24-hour off sedation” period would be reported, and no sedatives would be administered during the off-sedation period. However, completely avoiding sedatives or pain medications for patients on ECMO and mechanical ventilation is often impractical. We have all the data, encompassing information from both before and during the 24-hour off-sedation period. Analyzing this data is challenging due to its complexity, necessitating a thorough, manual review of charts to calculate each as-needed (PRN) dose and continuous infusion. Nevertheless, we have established stringent criteria for the ‘24-hour off sedation’ period to minimize confounding variables. Specifically, during this 24-hour timeframe: 1) no continuous sedative infusions are permitted, except for dexmedetomidine, and 2) the use of paralytics is prohibited, irrespective of dosage. These strict criteria may inadvertently exclude many potential participants (for example, those with a GCS of 15 while on a fentanyl infusion of 50ug/kg/hr). However, our primary objective is to ensure the selected patient group is as free from sedatives as possible. 6) We established a 24-hour off-sedation period as a threshold because 24 hours is a commonly accepted duration in the health care community to show stability of a medical condition³⁰. But the actual off-sedation duration for patients typically ranged from 3 to 5 days. Nonetheless, this ‘washout’ period might not suffice for patients who received higher sedative doses, were sedated for extended periods, experienced significant metabolic disturbances, or had large body mass indices (BMIs). 7) We have shown that the patients with coma_off had fewer (VV-ECMO) or even zero (VA-ECMO) good outcomes at discharge. Although the coma_off group was smaller which caused a higher number in the coma_on group, this does not change our conclusion that coma_off is a surrogate for poor neurologic outcome at discharge. But it is crucial to recognize a considerable fraction of patients with non-coma_off also had poor outcomes at discharge. Therefore, coma_off is a sensitive, but not specific indicator for poor outcome at hospital discharge.

Conclusions

Our findings highlight the importance of understanding the impact of coma on the outcome of ECMO patients. While proper sedation is crucial in the management of ECMO patients, our study underscores the necessity of implementing a standardized sedation weaning protocol. By doing so, we can more accurately assess the coma status, which may serve as a surrogate marker for poor outcomes upon hospital discharge.

HERALD Investigators:

Daniel Brodie, David Hager, Steven P. Keller, Errol L. Bush, R. Scott Stephens, Shivalika Khanduja, Jin Kook Kang, Ifeanyi David Chinedozi, Zachary Darby, Hannah J. Rando, Andrew Kalra, Trish Brown, Jiah Kim, Christopher Wilcox, Albert Leng, Andrew Geeza, Arjun Kumar Menta, Armaan F. Akbar, Benjamin L. Shou, David Zhao, Jaeho Hwang, Marc Sussman, Pedro Alejandro Mendez-Tellez, Philip Sun, Karlo Capili, Ramon Riojas, Diane Alejo, Scott Stephen, Harry Flaster