Population Characteristics and Markers of Withdrawal of Life Sustaining Therapy in Extracorporeal Membrane Oxygenation Therapy

Abstract

Objective:

As survival with extracorporeal membrane oxygenation (ECMO) therapy improves, it is important to study patients who do not survive in the context of withdrawal of life sustaining therapy (WLST). We sought to determine the population and clinical characteristics and to understand those who experienced short latency to WLST.

Design:

Retrospective cohort study

Setting:

Single academic hospital center

Participants:

Adult ECMO patients

Interventions:

None

Measurements and Main results:

150 patients (mean age 54.8±15.9, 43.3% female,) underwent ECMO (80% venoarterial-ECMO, and 20% venovenous-ECMO) during the study period. 73 (48.7%) patients had WLST from ECMO support at a median of 5 days; 33 (45.2%) patients had early WLST (≦5 day). Patients with WLST were older (49.6±14.7 vs 60.3±15.3, p<0.001), had elevated body mass index (28.3±5.5 vs. 31.7±7.6 kg/m²; p=0.002), longer ECMO duration (6 vs. 4 days; p=0.01), higher Acute Physiology and Chronic Health Evaluation (APACHE-II) (25 vs. 21, p<0.001) and Sequential Organ Failure Assessment (12 vs. 11, p=0.037) scores. Family request was frequently (91.7%) cited as part of WLST decision. WLST patients experienced more chaplaincy (89% vs. 65%, p<0.001) and palliative care consults (53.4% vs 29.9%, p=0.003), and code status change (Do Not Resuscitate: 83.6% vs. 7.8%, p<0.001).

Conclusions:

Our study demonstrates that nearly 50% ECMO patients underwent WLST, with approximately 25% occurring within the first 72 hours. These patients were older, greater illness severity and experienced a different clinical context. Neurological injury was not a primary reason for WLST in ECMO patients.

Introduction

As the overall use of extracorporeal membrane oxygenation (ECMO) and survival rises, it becomes increasingly important to understand the characteristics of patients who do not survive post-cannulation.¹ ECMO decannulation can be a form of withdrawal of life sustaining treatment (WLST) and WLST is often linked to perceived poor prognosis.^2,3 Neurological injury is a major factor in both in-hospital and long-term morbidity and mortality for ECMO patients, and is generally considered a poor prognostic factor.^4,5 However, there are limited data on the role of neurological injury in WLST of ECMO patients.^5,6 Furthermore, a paucity of literature exists regarding factors that are associated with WLST in ECMO. One prior study reports a survey of physician expert opinion on when to suggest limiting ECMO trials due factors such as age, baseline comorbidities, and additional organ failures.⁷ Others explore on the ethical issues pertaining to withdrawing ECMO, particularly when no further treatments are available.^7–10 In patients with other high mortality diseases, such as cardiac arrest and trauma, factors associated with WLST include neurologic injury, unwitnessed arrest, duration of arrest, race, female sex, and older age.^2,3,11–18 It is important to recognize and study these factors because they may influence the treating physician and that could lead to a decision of premature WLST.^2,18 Premature WLST has been recognized as a major issue in cardiac arrest and intracerebral hemorrhage.^2,11,12,18 As such, it is important to describe the timing, reasons leading to, and characteristics of WLST in ECMO patients.

In the only prior dedicated study of WLST following ECMO cannulation, a cohort of 235 patients demonstrated that approximately 53% of those who died experienced WLST requested by their families.⁸ The study also examined clinical factors that were associated with a care limitation pathway, including presence of advanced directive in the chart, the number of palliative care and chaplaincy consults, and change in code status.⁸ All of these were viewed as supportive initial steps in limiting aggressive medical care prior to a WLST.⁸ However, this study did not explore if other comorbidities such as neurological injury were associated with WLST or the timing of WLST in ECMO patients.

Herein, we aimed to describe the timing and characteristics of patients for whom ECMO support was withdrawn. We hypothesized that patients with WLST would demonstrate markers of more severe critical illness, more neurological complications, and experience more clinical factors that may be associated with a clinical context of limiting care. Furthermore, we aimed to compare risk factors associated with short latency to WLST in patients with ECMO support.

Methods

Study Design and Feasibility

We conducted a retrospective observational study of consecutive patients with ECMO support who met study criteria previously described in Cho et al. 2019.⁴ Briefly, we included all adult patients (age >18 years) who received venoarterial (VA-) or venovenous (VV-) ECMO. The study was carried out at a tertiary care center with the approval of the local institutional review board. Patients from November 2017 to August 2019 were included, as those were the charts completely abstracted at time of study analysis. Indications for VA-ECMO at our institution were: cardiogenic shock from reversible cause, inability to come off cardiopulmonary bypass post-surgery, bridge to transplant, bridge to ventricular assist device, bridge to decision, massive pulmonary embolus, severe pulmonary hypertension with right heart failure, septic shock with cardiomyopathy, and extracorporeal cardiopulmonary resuscitation (E-CPR). Indications for VV-ECMO at our institution were: patients with persistent PO₂< 60 mmHg, PCO₂ > 60 mmHg with acidosis, or requiring paralysis for > 24 hours for respiratory failure following use of maximal and advanced ventilator strategies. Exclusion or contraindications for ECMO cannulation included age > 65 years, severe and irreversible neurologic injury, cardiopulmonary resuscitation > 60 minutes, chronic end organ dysfunction, malignancy with life expectancy less than five years, severe and irreversible coagulopathy, inability to anticoagulate, severe immunocompromised state, severe peripheral vascular disease, multiorgan failure, invasive infection with low likelihood of resolution or survival, aortic insufficiency, or aortic dissection. Patients were cannulated by one of a team of several surgeons with expertise in ECMO cannulation. Consent was obtained as part of surgical consent process for all post-cardiotomy patients; for other patient’s consent was obtained if possible. A notable exception was for E-CPR patients as they were all deemed emergent. ECMO patients were managed in either the medical intensive care unit, cardiovascular surgery intensive care unit, or cardiac intensive care unit supervised by a team of attending intensivists. ECMO circuit was co-managed by a team of perfusionists who checked the circuit every two hours.

All patients were reviewed from day one of ECMO cannulation to hospital discharge or death. A baseline neurologic exam on day one, Glasgow Coma Scale (GCS) score, Acute Physiology and Chronic Health Evaluation II (APACHE II), and Sepsis-related Organ Failure Assessment (SOFA) scores were gathered. Charts were reviewed for demographic data; ECMO mode and cannulation; ECMO indication; WLST with ECMO decannulation; reason for WLST based on nursing or physician notes; if patients underwent chaplain, palliative care, or ethics consultations while on ECMO; documentation of a family meeting while on ECMO; change of code status (Do Not Resuscitate [DNR] and/or Do Not Intubate [DNI]) prior to withdrawal of ECMO; and/or presence of an advance directive in the chart. WLST with ECMO decannulation was determined based on nursing, physician, family meeting, or discharge notes with specific mention of withdrawal and no other reason (brain death, futility, asystole) listed. Chaplaincy, palliative care, or ethics consults, family meetings, changes in code status, or new advanced directives charted after decannulation were not included. Prior medical history and reasons for WLST were not mutually exclusive (meaning more than one option could be chosen) in order to better reflect patient complexity.

Statistical Analysis

Comparison of demographic and clinical variables in patients with WLST versus all other patients who did not undergo WLST included in the study were performed using an unpaired independent sample T-test, Mann-Whitney U, or Chi Square analysis as appropriate. Secondarily, a skew analysis was performed to determine early versus late WLST, defined as WLST occurring before or after 5 days after ECMO cannulation, respectively. Results are expressed as the mean with standard deviation if normally distributed or as the median with quartiles if non-normally distributed for quantitative variables and as proportions for categorical variables. Nominal logistic multivariate analysis was performed on the most significantly associated variables with odds ratios and confidence intervals reported. A p-value of less than 0.05 was considered statistically significant. All analyses were performed using IBM SPSS Statistics 25 (Armonk, New York, USA).

Results

Study Population

A total of 150 patients (mean age 54.8 ± 15.9 years) were cannulated during the 22-month study period. Baseline characteristics and clinical factors are summarized in Table 1. The most common past medical history included hypertension (51.3%), hyperlipidemia (51.3%), and prior antiplatelet therapy use (46%) (Supplemental Table 1). Of the 120 VA-ECMO patients, the indications were cardiogenic shock (N=60, 50%), post-cardiotomy shock (N=52, 43.3%), and E-CPR (N=8, 6.67%) (Supplemental Table 2). Ninety-six (64%) patients supported by ECMO died during their index hospitalization (Table 1).

Table 1.

Comparison of Withdrawal of Life Sustaining Therapy to No Withdrawal of Life Sustaining Therapy.

	Total (N=150)	No WLST (N=77)	WLST (N=73)	p
Age Mean ± SD	54.8 ± 15.9	49.6 ± 14.7	60.3 ± 15.3	<0.001
BMI (Kg/m2) Mean ± SD	30.0 ± 6.8	28.3 ± 5.5	31.7 ± 7.6	0.002
Sex (Female) N (%)	65 (43.3)	36 (46.8)	29 (39.7)	0.385
Duration of ECMO (Days) Median [IQR]	5 [2.5, 10]	4 [2,7]	6 [3.5,11]	0.010
Race N (%)				0.324
Asian	10 (6.7)	4 (5.19)	6 (8.2)
Black	36 (24)	24 (31.1)	12 (16.4)
Hispanic	6 (4.0)	3 (3.90)	3 (4.1)
White	88 (58.7)	41 (53.2)	47 (64.4)
Other	9 (6.0)	5 (6.5)	4 (5.5)
ECMO Mode N (%)				0.002
VA	120 (80)	54 (70.1)	66 (90.4)
VV	30 (20)	23 (29.9)	7 (9.6)
Cannulation N (%)				0.036
Central	65 (43.3)	27 (35.1)	38 (52.1)
Peripheral	85 (56.7)	50 (64.9)	35 (47.9)
Death N (%)	96 (64.0)	23 (29.9)	73 (100)
Pre-ECMO GCS Median [IQR]	15 [3,15]	15 [10, 15]	15 [6.75, 15]	0.121
Baseline pH Mean ± SD	7.26 ± 0.14	7.26 ± 0.14	7.25 ± 0.14	0.944
APACHE II Score Day 1 Median [IQR]	23 [18, 28]	21 [16,25]	25 [22, 29]	<0.001
SOFA Score Day 1 Median [IQR]	11 [10, 13]	11 [9,12.75]	12 [10, 14]	0.037
Pump Flow Rate at 24 hours L/Minute ±SD	4.15 ± 0.98	4.10 ± 0.92	4.22 ± 1.1	0.492
Clinical Markers N (%)
Chaplaincy Consult	115 (76.7)	50 (65.0)	65 (89.0)	<0.001
Palliative Care Consult	62 (41.30	23 (29.9)	39 (53.4)	0.003
Ethics Consult	1 (0.67)	0 (0)	1 (1.4)	0.303
Family Meeting Documented	91 (60.7)	25 (32.5)	66 (90.4)	<0.001
Advance Directive	22 (14.7)	11 (14.3)	11 (15.1)	0.892
New DNR Order	67 (44.7)	6 (7.8)	61 (83.6)	<0.001
New DNI Order	11 (7.33)	0 (0)	11 (15.1)	<0.001

Note: WLST = Withdrawal of Life Sustaining Therapy. BMI = Body Mass Index. ECMO = Extracorporeal Membrane Oxygen Therapy.VA = Venoarterial. VV = Venovenous. GCS = Glasgow Coma Score. APACHE II= Acute Physiology and Chronic Health Evaluation. SOFA = Sequential Organ Failure Assessment. DNR = Do Not Resuscitate. DNI = Do Not Intubate.

Withdrawal of ECMO

Of the 96 patients who died, 73 (76.0%) patients died in the context of WLST from ECMO (Table 1). None of the patients in the WLST group survived. The other 23 patients died either while on ECMO without WLST (N=12, 12.5%) or after ECMO decannulation during the index hospitalization (N=11, 11.5%). Nine (75%) of the 12 patients who died without WLST on ECMO were declared brain dead. ECMO duration was significantly longer for those who underwent WLST than no WLST (median 6 vs. 4 days, p=0.01) (Table 1). Patients who experienced WLST were significantly older (49.6±14.7 vs. 60.3±15.3, p<0.001) (Table 1). Thirty-three (45.2%) patients in the WLST group were withdrawn less than 5 days from ECMO cannulation (Figure 1). VA-ECMO patients (90.4%) more frequently underwent WLST than VV-ECMO patients (9.6%, p=0.002) (Table 1). Similarly, patients with central cannulation (52.1%) were also more likely to undergo WLST than those with peripheral cannulation (47.9%; p= 0.036) (Table 1). APACHE II and SOFA (Day 1) scores were higher in the WLST group (APACHE II: 25 vs. 21, p<0.001; SOFA: 12 vs. 11, p=0.037) (Table 1). There was no difference for WLST by VA-ECMO indication (Supplemental Table 2). In a multivariable logistic regression analysis age (adjusted odds ratio [aOR]=1.06, 95% CI 1.03–1.09, p <0.0001), BMI (aOR=1.08, 95% 1.02–1.15, p=0.0094), ECMO duration (aOR=1.00, 95% CI 1.00–1.01, p=0.0004), VA-ECMO (aOR=7.61, 95% CI 2.03–28.53, p=0.0009), and APACHE II (aOR=1.08, 95% CI 1.02–1.15, p=0.0037) were independently associated with WLST (Table 4).

Figure 1.

Histogram of withdrawal of life sustaining therapy by day.

Table 4.

Multivariable Logistic Regression Analysis of Factors Associated with WLST

	Univariate Logistic Regression		Multivariate Logistic Regression
	Odds Ratio (95% CI)	p	Odds Ratio (95% CI)	p
Baseline Characteristic
Age	1.06 (1.02–1.09)	<0.0001	1.06 (1.03–1.09)	<0.0001
BMI	1.08 (1.01–1.15)	0.013	1.08 (1.02–1.15)	0.0094
Hours of ECMO	1.00 (1.00–1.01)	<0.001	1.00 (1.00–1.01)	0.0004
Type of ECMO
VA ECMO	7.70 (1.96–30.3)	0.004	7.61 (2.03–28.53)	0.0009
VV ECMO	0.13 (0.03–0.51)	0.004	0.13 (0.035–0.50)	0.0009
Cannulation
Central	1.16 (0.49–2.76)	0.735
Peripheral	0.86 (0.36–2.05)	0.735
Illness Severity Markers
APACHE II Score	1.07 (1.00–1.13)	0.037	1.08 (1.02–1.15)	0.0037
SOFA Score	1.10 (0.94–1.28)	0.226

Abbreviations: WLST= withdrawal of Life sustaining therapy; CI= confidence interval, BMI: body mass index, ECMO= extracorporeal membrane oxygen therapy, APACHE II= Acute Physiology and Chronic Health Evaluation. SOFA = Sequential Organ Failure Assessment.

Of the non-mutually exclusive reasons for WLST, the most commonly cited was family request (91.7%), followed by multi-organ failure (48.0%), and primary organ damage determined to be non-survivable or non-operable (41.1%), as summarized in Table 2. Isolated neurologic reasons for WLST such as coma (8.3%), ischemic stroke (8.3%), or hemorrhagic stroke (1.4%) were not frequently observed (Table 2). Clinical markers such as chaplaincy consult (89% vs. 65% p<0.001), palliative care consult, (53.4% vs. 29.9%, p=0.003), documented family meeting (90.4% vs. 32.5%, p<0.001), new DNR order placed (83.6% vs. 7.8%, p<0.001), and new DNI order (15.1% vs. 0%, p<0.001) were more commonly observed in patients who underwent WLST vs. no WLST (Table 1).

Table 2.

Reasons for Withdrawal of Life Sustaining Therapy

Reason for Withdrawal	Frequency (N=73) N (%)	Early WLST (N=33) N (%)	Late WLST (N=40) N (%)
Family Request	67 (91.7)	29 (87.9)	38 (95)
Multi-Organ Failure	35 (48.0)	12 (36.4)	23 (57.5)
Primary Organ Damage Non-Survivable or Operable	30 (41.1)	14 (42.4)	16 (40)
Non-CNS Hemorrhages	12 (16.4)	8 (24.2)	4 (10)
Coma	6 (8.3)	4 (12.1)	2 (5)
Ischemic Stroke	6 (8.3)	4 (12.1)	2 (5)
Hemorrhagic Stroke	1 (1.4)	1 (3.0)	0 (0)
Other	5 (6.8)	2 (6.1)	3 (7.5)

Note: WLST = Withdrawal of Life Sustaining Therapy, CNS = Central Nervous System.

Early Compared to Late WLST

WLST occurred at a median of 5 days (IQR=3.5–11), which was used to group early versus late WLST in our study. Baseline characteristics comparing early to late WLST are summarized in Table 3. Late WLST was more frequent in patients with Black race (p=0.0035) compared to other races (Table 3). Type of ECMO mode also significantly differed with all of the VV-ECMO patients undergoing late WLST (17.5%) (p=0.011) (Table 3). The most common (non-mutually exclusive) reasons for early WLST were family request (83.3%), primary organ damage determined to be non-survivable or non-operable (38.9%), and multiorgan failure (36.1%) (Table 2). Early WLST had few neurologic reasons cited including coma (12.1%), ischemic stroke (12.1%), and hemorrhagic stroke (3%) (Table 2). The top three reasons cited by late WLST were similar: family request (95%), multiorgan failure (57.5%), primary organ damage determined to be non-survivable or non-operable (40%) (Table 2). Late WLST had even fewer neurologic reasons cited: coma (5%) and ischemic stroke (5%). Only palliative care consults had increased prevalence as a marker of WLST between early and late WLST (72.5 % vs. 30.3 %, p<0.001) (Table 3).

Table 3.

Comparing Early Versus Late Withdrawal of Life Sustaining Therapy

	Early WLST (N =33)	Late WLST (N=40)	p
Age Mean ± SD	63.6 ± 14.6	57.6 ± 15.58	0.687
BMI (Kg/m2) Mean ± SD	32.3 ± 7.7	31.2 ± 7.6	0.934
Sex (Female) N (%)	11 (33.3)	18 (45)	0.311
Duration (Days) Mean ± SD	2.9 ± 1.52	13.00 ± 9.58	0.002
Race N (%)			0.035
Asian	5 (15.2)	1 (2.5)
Black	2 (6.1)	10 (25)
Hispanic	3 (9.1)	0 (0)
White	21 (63.6)	26 (65)
Other	2 (6.1)	2 (5)
ECMO Mode N (%)			0.011
VA	33 (100)	33 (82.5)
VV	0 (0)	7 (17.5)
Cannulation N (%)			0.184
Central	20 (60.6)	18 (45)
Peripheral	13 (39.4)	22 (55)
Pre-ECMO GCS, Median [IQR]	15 [6, 15]	15 [9, 15]	0.809
Baseline pH Mean ± SD	7.24 ± 0.14	7.25 ± 0.13	0.504
APACHE II Score Day 1 Median [IQR]	25.5 [23, 31]	23 [19.75, 28.25]	0.082
SOFA Score Day 1 Median [IQR]	11.5 [10, 14]	12 [10, 14.25]	0.462
Pump Flow Rate at 24 hours L/Minute ±SD	3.95 ± 1.19	4.34 ± 0.91	0.094
Clinical Marker N (%)
Chaplaincy Consult	28 (84.8)	37 (92.5)	0.298
Palliative Care Consult	10 (30.3)	29 (72.5)	0.000
Ethics Consult	0 (0)	1 (2.5)	0.360
Family Meeting Documented	28 (84.8)	38 (95)	0.143
Advance Directive	4 (12.1)	7 (17.5)	0.523
New DNR Order	27 (81.8)	34 (85)	0.715
New DNI Order	4 (12.1)	7 (17.5)	0.523

Note: WLST = Withdrawal of Life Sustaining Therapy. BMI = Body Mass Index. ECMO = Extracorporeal Membrane Oxygen Therapy.VA = Venoarterial. VV = Venovenous. GCS = Glasgow Coma Score. APACHE II= Acute Physiology and Chronic Health Evaluation. SOFA = Sequential Organ Failure Assessment. DNR = Do Not Resuscitate. DNI = Do Not Intubate.

Discussion

Our study has several important findings. First, our study demonstrated that most deaths (76%) in our cohort were related to WLST. This is consistent with the prior study of WLST in a single center ECMO cohort reporting 53% of deaths due to WLST.⁸ The main reason for non-WLST deaths in our cohort were brain deaths. However, for those with WLST, neurological injury was not a major factor in the withdrawal decision in contrary to our hypothesis. In fact, neurological injury was only reported as a reason for WLST in 18% of our cohort, which is different than what’s reported in cardiac arrest.^2,3,11,18 The most common reason for WLST was family request (91.7%), followed by multi-organ failure (48%), and primary organ damage deemed non-operable or non-survivable (41.1%), consistent with a prior paper on physician’s views of when ECMO should be limited.⁷ This suggests that medical care was withdrawn primarily because of underlying refractory cardiopulmonary failure requiring ECMO cannulation rather than neurological injury. Although our patients had a much higher percentage of family request cited as a reason for WLST (91.7%) compared to another ECMO cohort with 26%,⁸ this finding may be due to different methodology, reporting, or institutional practice.

Secondly, our study demonstrated that sicker patients with longer ECMO runs were more likely to undergo WLST, as indicated by higher APACHE II and SOFA scores. Similarly, centrally cannulated patients, which is generally associated with higher mortality and illness severity than peripherally cannulated, patients were more likely to undergo WLST.¹⁹ Age has been previously associated with WLST, which was consistent in our study.^14,17 We did not observe any sex-related differences, unlike prior cardiac arrest studies where WLST was associated with female sex.^14,17 Although race/ethnicity did not change the frequency of WLST, we found that Black race was associated with late WLST. Since it is well known that Black and Hispanic races are less likely to experience WLST,¹⁵ further research on equity of ECMO care with race/ethnicity as well as sex/gender is warranted.

Thirdly, we demonstrated that early WLST was very common with 45% patients undergoing early WLST. Interestingly, nearly 25% of all WLST occurred in the first 72 hours of ECMO support (Figure 1), which may be premature to prognosticate and predict outcomes. The early WLST group had more neurological injury than those with late WLST, but is important to note that overall frequency of neurological injury in this cohort was relatively low. At this time, early neurological prognostication within first 72 hours is currently not recommended and should be avoided.^2,18,20,21 Currently there is little guidance on “adequate” duration of ECMO support before clinicians decide that further medical care is futile.⁷ Although several studies have evaluated factors relating ECMO mortality,^22–24 little emphasis has been placed on WLST, a major factor for mortality. There are no prediction scores that utilizes clinical and laboratory variables during ECMO support to predict patient outcome in ECMO. Furthermore, RESP and SAVE scores do not account for mortality due to WLST, which is a major limitation for these prediction scores^25,26. Identifying and understanding on-ECMO factors may influence a provider’s perspective on an ECMO patient’s prognosis. Extrapolating from cardiac arrest literature and our study, which is primarily based on VA-ECMO patients, with the limitations of current ECMO prognostication, we advise that clinicians to avoid prognosticating and limit WLST in the first 72 hours for VA ECMO patients.

Finally, the clinical context a patient experiences prior to WLST may be representative of their overall outcome. At our institution, patients on ECMO did not automatically have chaplaincy or palliative care consults and there is no standard protocol for documentation of family meetings. Patients undergoing WLST were more likely to have chaplaincy (89 vs. 65%) and palliative care consults (53.4 vs. 29.9%), change in DNR (83.6 vs. 7.8%) and/or DNI (15.1 vs. 0%) orders placed during their ECMO run. In comparison to a prior study, we observed an overall lower frequency of advance directives in the WLST group (15% vs. 37%), but similar or higher rates of palliative care (53.4% vs. 16%), chaplaincy consults (89% vs. 94%), and DNR order placement (83.6% vs. 29%).⁸ The occurrence of these events likely represents the overall perceived illness severity and prognosis for the patient by the provider team. Early versus late WLST did not differ significantly with regards to clinical events except palliative care consults (30.3% vs. 72.5% respectively), likely reflecting an opportunity for the completion of palliative consultation. This is consistent with an earlier finding that patients who received palliative care consults had significantly longer ECMO runs.⁶ These consults and events may be associated with WLST because they are steps along the pathway to WLST. These consults may support the family and provider team when a poor outcome is expected. A standardized institutional protocol utilizing palliative care, addressing advance directives, chaplaincy support, and early family involvement when older patients with elevated APACHE II and SOFA scores may improve communication and critical care of ECMO patients and their families.^27,28

Our study has several limitations. This study is a single-center retrospective review. Studies using large ECMO databases are necessary in understanding the population characteristics of patients with ECMO that undergo WLST. Furthermore, a prospective evaluation could help address a direct relationship with regards to the clinical environment and WLST. This study cohort combined VA- and VV-ECMO patients due to small sample size. All VV-ECMO patients were observed in the late WLST group as the average ECMO days are longer on in patients with VV-ECMO. Our findings are therefore more representative of a VA-ECMO population. In the future, further VA-ECMO subgroup analysis of the post-cardiotomy and E-CPR patients would be valuable as they are distinct in the ECMO population. Timing of WLST may be different based on ECMO indications. Our study was not able to address this question separately due to the small sample size. However, our study represents one of the largest studies to describe the characterization of WLST in ECMO. A prospective study on the latency to WLST with a protocolized end-of-life care approach will enhance understanding of risk factors, timing, and reasons for WLST in ECMO patients. We were unable to evaluate which additional organ support (such as ventilator or renal replacement therapy) was crucial when multiorgan failure was cited as a reason for WLST. Qualitative studies with provider teams and health care decision makers may better address the granularity of these complex decisions and the weight of particular factors. Further studies on this important topic are warranted to develop a predictive model and prognostication algorithm to guide clinicians in end-of-life care of ECMO patients.

Conclusions

Our study demonstrated that the majority of ECMO patients who died underwent WLST and approximately 25% of those had early WLST within 72 hours and neurological injury was not a primary reason for WLST in ECMO patients. Patients with advanced age, higher SOFA/APACHE II scores, and central cannulation were more likely to have WLST. Future research with a protocolized multidisciplinary end-of-life management is necessary to improve overall care.