Early Low Pulse Pressure in VA-ECMO is Associated with Acute Brain Injury

Abstract

Background

Pulse pressure is a dynamic marker of cardiovascular function and is often impaired in patients on veno-arterial extracorporeal membrane oxygenation (VA-ECMO). Pulsatile blood flow also serves as a regulator of vascular endothelium, and continuous flow mechanical circulatory support can lead to endothelial dysfunction. We explored the impact of early “low” pulse pressure on occurrence of acute brain injury (ABI) in VA-ECMO.

Methods

We conducted a retrospective analysis of adults with VA-ECMO at a tertiary care center between July/2016-January/2021. Patients underwent standardized multimodal neuromonitoring throughout ECMO support. ABI included intracranial hemorrhage, ischemic stroke, hypoxic ischemic brain injury, cerebral edema, seizure, and brain death. Blood pressures were recorded every 15 minutes. Low pulse pressure was defined as a median pulse pressure <20 mmHg in the first 12 hours of ECMO. Multivariable logistic regression was performed to investigate the association between pulse pressure and ABI.

Results

We analyzed 5,138 blood pressure measurements from 123 (median age=63, 63% male) VA-ECMO patients (54% peripheral; 46% central cannulation), of which 41 (33%) experienced ABI. Individual ABI rates were: ischemic stroke (n=18, 15%), hypoxic ischemic brain injury (n=14, 11%), seizure (n=8, 7%), intracranial hemorrhage (n=7, 6%), cerebral edema (n=7, 6%), and brain death (n=2, 2%). Fifty-eight (47%) patients had low pulse pressure. In a multivariable model adjusting for pre-selected covariates including cannulation strategy (central vs. peripheral), lactate on ECMO day 1, and left ventricle venting strategy, low pulse pressure was independently associated with ABI (adjusted Odds Ratio=2.57, 95% Confidence Interval=1.05-6.24). In a model with the same covariates, every 10-mmHg decrease in pulse pressure was associated with a 31% increased odds of ABI (95%CI: 1.01-1.68). In a sensitivity analysis model adjusting for systolic pressure, pulse pressure remained significantly associated with ABI.

Conclusions

Early low pulse pressure (<20 mmHg) was associated with ABI in VA-ECMO patients. Low pulse pressure may serve as a marker of ABI risk, which necessitates close neuromonitoring for early detection.

Introduction

Venoarterial extracorporeal membrane oxygenation (VA-ECMO) is an increasingly popular mode of mechanical circulatory support for patients with refractory heart failure¹. However, VA-ECMO patients still have a significant mortality rate, driven in part by a high incidence of acute brain injury (ABI). Extracorporeal Life Support Organization (ELSO) Registry studies suggest ABI occurs in 11-20% of adult VA-ECMO patients^2,3, although the true incidence may be even higher when accounting for sub-clinical ABI. In our institution, we have previously reported a higher ABI frequency of 33% when patients were assessed with standardized neuromonitoring^4,5. Several physiologic risk factors for ABI have been identified including early hyperoxia, rapid correction of peri-cannulation PaCO₂, shortened duration of early hypothermia, and total duration of ECMO support^2,6-8.

Arterial pulse pressure has been associated with cardiovascular outcomes in the general population and those with pre-existing heart failure or atherothrombotic disease^9-11. Pulse pressure tends to increase with advanced age as compliance of the large arteries decreases¹². High pulse pressure is associated with major adverse cardiac events⁹ while very low pulse pressure is a marker of worse underlying ventricular function¹⁰. In the setting of mechanical circulatory support, low pulse pressure in the first 24 hours of extracorporeal cardiopulmonary resuscitation has been associated with lower probability of weaning from ECMO^13,14.

VA-ECMO provides non-pulsatile flow to patients with already severely compromised cardiac function. Non-pulsatile flow disrupts several physiological systems through a complex and incompletely understood mechanism surrounding endothelial dysregulation^15,16. These vascular processes may predispose patients to ABI, such as ischemic or hemorrhagic strokes, similar to patients with left ventricular assist device with non-physiological, non-pulsatile cerebral blood flow¹⁷. Additionally, pulse pressure may be modulated during VA-ECMO through adjusting pump speed and the use of additional mechanical circulatory support devices such as intra-aortic balloon pump (IABP) and Impella, or with inotropic agents. Thus, pulse pressure could be a valuable and potentially modifiable risk factor in the prediction of ABI. We hypothesized that low pulse pressure while on VA-ECMO is associated with an increased risk of ABI. Additionally, we conducted an exploratory analysis of the impact of blood pressure variability (BPV) on ABI as BPV has been shown to be a prognostic marker in non-ECMO populations^18,19.

Methods

Study Design

This study was approved by the Johns Hopkins Hospital Institutional Review Board (IRB00216321) on 10/22/2019. Informed consent was waived as this was an observational study. We retrospectively analyzed patients with VA-ECMO support at a tertiary center between July 2016 and January 2021. This study derives from a multidisciplinary effort and an initiative between the Cardiovascular Intensive Care Unit, Cardiac Critical Care Unit, and Neuroscience Critical Care Unit to improve overall clinical care and outcomes for patients with ECMO support. All ECMO patients received routine neurocritical care consultations and standardized neuromonitoring per our institutional protocol^4,5.

Neuromonitoring

Our neuromonitoring protocol consists of neurocritical care consultations with serial neurologic examinations, transcranial doppler, electroencephalography, somatosensory evoked potentials, and neuroimaging including computed tomography (CT) during ECMO and magnetic resonance imaging before discharge⁴. Clinically significant neurological diagnoses are confirmed using CT brain during the ECMO run.

Participants

All adult (≥18 years old) patients who were supported with VA-ECMO were included in this study. We excluded patients who required intraoperative ECMO support only.

Data Collection

We collected baseline characteristics including demographics, social history, past medical history, baseline neurological, cardiac, and pulmonary diagnoses. Additionally, ECMO characteristics such as cannulation site and use of left ventricle venting strategy were collected. Laboratory and physiologic values were extracted from the electronic medical record (EMR). Serial blood pressure (BP) measurements from the bedside monitors were automatically recorded and linked to the EMR flowsheet every 15 minutes during first 12 hours following ECMO cannulation.

Definitions

We analyzed arterial line BP collected during the 12 hours following cannulation due to the consistent granularity of the data across all patients. Low pulse pressure was defined as a median pulse pressure <20 mmHg. We chose to dichotomize pulse pressure according to an a priori 20-mmHg cutoff since it provides a reasonable potential clinical target, however, we conducted a sensitivity analysis at 15 and 25 mmHg cutoffs which supports the robustness of our model. Additionally, since we had a large amount of serial blood pressure measurements per patient, we calculated the duration of time each patient spent with a pulse pressure <20 mmHg in the first 12 hours of ECMO. Since BPV has been associated with poor functional neurologic outcomes in stroke and ICU patients^20,21, we conducted an exploratory analysis of BPV on ABI. We calculated BPV for systolic and diastolic pressure in the first 12 hours after ECMO initiation using two generally accepted methods: standard deviation (SD) and coefficient of variation (CV)²². Arterial blood gasses and lactate were recorded as the nadir value of all collected measurements in the first 12 hours of ECMO support. ABI was defined as ischemic stroke, intracranial hemorrhage (ICH), hypoxic ischemic brain injury (HIBI), cerebral edema, seizure, or brain death. Individual ABI diagnoses were not mutually exclusive (a patient could have more than one ABI diagnosis). Functional neurological outcome was determined by the modified Rankin scale (mRS) at discharge; good and poor neurologic outcomes were defined as mRS 0-3 and 4-6, respectively. Left ventricle (LV) venting strategies included IABP, percutaneous ventricular assist device (i.e., Impella), right superior pulmonary vein and left ventricle apex vents.

Outcomes

Our primary outcome was the presence of ABI during VA-ECMO support. A secondary outcome was in-hospital mortality.

Statistical Analysis

Data were presented as median (interquartile range: IQR) for continuous variables and absolute counts with percentages for binary and categorical variables. Demographic and clinical characteristics in patients with and without events were compared using Wilcoxon rank-sum test for continuous variables and Pearson’s chi-square test for binary or categorical variables. Odds ratios (ORs) and adjusted ORs (aORs) with 95% confidence intervals (CIs) were calculated by multivariable logistic regression models fitted by penalized maximum likelihood regression to account for a low exposure rate for some covariates. We selected covariates a priori based on known risk factors for development of ABI and the best performing model was chosen by lowest Akaike’s information criterion. Worse performing models with alternative covariates are provided in the Supplement. A p value less than 0.05 was considered statistically significant. In an exploratory analysis, pulse pressure was treated as a continuous variable. The relationship between pulse pressure and systolic BP was analyzed via non-linear regression with a 3-parameter logistic function. The relationship between pulse pressure and mean arterial pressure (MAP) was evaluated with standard linear regression.

Although our model adjusted for confounding variables, we separately analyzed three sub-populations: those with LV venting, those with post-cardiotomy shock (PCS), and those with non-PCS cardiogenic shock, as these patients may have confounding effects on outcome. More specifically, LV venting can modify pulsatility, and PCS patients tend to have worse outcome relative to other VA-ECMO indications²³. Finally, we conducted an exploratory analysis by making separate models for ischemic stroke and hemorrhagic stroke.

All statistical analyses were performed using STATA 17 (StataCorp, College Station, TX, USA).

Results

Out of a total 129 VA-ECMO patients, we included 123 (median age=63, 63% male) who met the inclusion criteria and had complete granular BP data. Six out of 129 patients were excluded due to irregular and sparse BP recordings. A total of 5,138 BP measurements were recorded during the first 12 hours after ECMO initiation, with a bimodal distribution (Figure 1A, S1). Central cannulation was performed in 57 (46%) patients while the remaining 66 (54%) were peripherally cannulated. Fifty-four (44%) patients had an LV vent, including 35 (28%) with IABP, 10 (8%) with Impella, 7 (6%) with direct pulmonary vein vent and 2 (2%) with direct LV apex vent.

Figure 1.

(A) Histogram of pulse pressure across 5,138 blood pressure measurements recorded every 15 minutes for the first 12 hours following ECMO initiation, for patients with (“ABI+”, in red) and without (“ABI−”, in green) ABI. (B) Patients with ABI had a lower pulse pressure than those without. ABI: acute brain injury.

Forty-one (33%) patients had ABI. The breakdown of ABI are as follows: ischemic stroke (n=18, 15%), hypoxic ischemic brain injury (n=14, 11%), seizure (n=8, 7%), intracranial hemorrhage (n=7, 6%), cerebral edema (n=7, 6%), and brain death (n=2, 2%). Demographics, baseline clinical characteristics including ejection fraction, and pre-ECMO severity of illness were similar between those with and without ABI, except that those with ABI had a greater proportion of prior ICH (5% vs. 0%, p=0.04). Arterial blood gases and lactate levels were also not different between the two groups. Those with ABI tended to be centrally cannulated (vs. peripherally, 66% vs. 48%, p=0.06). There were no differences in LV venting practices between those with and without ABI (Table 1). IABP patients, who had a higher pulse pressure (46 vs. 14 mmHg, p<0.001), did not experience less ABI compared to others (31% vs. 34%, p=0.77).

Table 1.

Baseline characteristics and clinical variables in patients with versus without ABI. Data are presented as median (IQR) for continuous variables and n (%) for categorical variables.

Variable	Total (n=123)	Patients without ABI (n=82, 67%)	Patients with ABI (n=41, 33%)	P value
Demographics
Age, years	59 (49-69)	62 (50-69)	55 (48-68)	0.30
Male	77 (63%)	54 (66%)	23 (56%)	0.29
Body Mass Index, kg/m2	30 (25-35)	30 (25-36)	29 (25-33)	0.45
Race				0.74
White	78 (63%)	54 (66%)	24 (59%)
Black	26 (21%)	15 (18%)	11 (27%)
Hispanic	2 (2%)	1 (1%)	1 (2%)
Asian	9 (7%)	7 (9%)	2 (5%)
Others	8 (7%)	5 (6%)	3 (7%)
Past medical history
Ischemic stroke	10 (8%)	6 (7%)	4 (10%)	0.64
Intracranial hemorrhage	2 (2%)	0 (0%)	2 (5%)	0.04
Hypertension	91 (74%)	59 (72%)	32 (78%)	0.47
Hyperlipidemia	68 (55%)	45 (55%)	23 (56%)	0.90
Heart failure	42 (34%)	29 (35%)	13 (32%)	0.69
Chronic kidney disease	18 (15%)	11 (13%)	7 (17%)	0.59
Atrial fibrillation	38 (31%)	26 (32%)	12 (29%)	0.78
Antiplatelet therapy before index hospitalization	64 (52%)	46 (56%)	18 (44%)	0.20
Anticoagulation before index hospitalization	29 (24%)	20 (24%)	9 (22%)	0.76
Baseline ejection fraction *				0.97
≤40%	62 (50%)	41 (50%)	21 (51%)
41-49%	13 (11%)	9 (11%)	4 (10%)
≥50%	44 (36%)	29 (35%)	15 (37%)
ECMO day 1 variables **
Glasgow Coma Scale	15 (5-15)	15 (6-15)	14 (3-15)	0.47
Cardiac arrest	53 (44%)	32 (40%)	21 (52%)	0.19
Inotrope or vasopressor support	100 (83%)	69 (86%)	31 (78%)	0.23
SOFA score	11 (10-13)	12 (10-13)	11 (10-14)	0.53
Lactate (mmol/L)	6.0 (2.8-10.1)	5.4 (2.6-10.0)	7.3 (3.5-13.6)	0.10
pH	7.27 (7.19-7.34)	7.29 (7.20-7.34)	7.25 (7.17-7.33)	0.27
PaO2 (mmHg)	99 (95-100)	99 (95-100)	99 (94-100)	0.44
PaCO2 (mmHg)	40 (34-46)	40 (34-46)	42 (34-50)	0.29
HCO3− (mEq/L)	19.0 (15.5-22.0)	19.0 (15.0-22.0)	19.0 (16.0-22.0)	0.91
On-ECMO blood pressure (mmHg)
Systolic BP	94 (82-104)	96 (86-108)	92 (80-100)	0.04
Diastolic BP	66 (54-74)	64 (52-74)	66 (59-74)	0.36
Mean arterial pressure	74 (67-83)	74 (66-85)	74 (67-82)	0.85
Pulse pressure	27 (10-44)	32 (10-46)	20 (8-29)	0.04
Low pulse pressure (<20 mmHg)	58 (47%)	33 (40%)	25 (61%)	0.03
Systolic BPV, SD method	12.0 (9.6-18.7)	11.9 (9.0-16.6)	12.0 (10.2-20.3)	0.34
Diastolic BPV, SD method	8.5 (6.0-12.0)	8.1 (5.6-11.5)	9.7 (7.4-12.3)	0.05
Systolic BPV, CV method	57.4 (29.8-104.5)	48.5 (26.8-94.7)	76.2 (40.5-141.8)	0.02
Diastolic BPV, CV method	32.7 (16.8-96.0)	30.4 (15.1-88.4)	61.3 (24.6-108.2)	0.03
ECMO indications
Cardiogenic shock	72 (59%)	49 (60%)	23 (56%)	0.70
Post-cardiotomy shock	51 (41%)	32 (39%)	19 (46%)	0.44
ECPR	23 (19%)	13 (16%)	10 (24%)	0.25
Left ventricular venting strategy				0.33
None	69 (56%)	42 (51%)	27 (66%)
IABP	35 (28%)	24 (29%)	11 (27%)
Impella	10 (8%)	9 (11%)	1 (2%)
Right superior pulmonary vein	7 (6%)	5 (6%)	2 (5%)
Left ventricle apex	2 (2%)	2 (2%)	0 (0%)
Intracardiac thrombus	5 (4%)	3 (4%)	2 (5%)	0.75
Cannulation strategy				0.06
Central	57 (46%)	43 (52%)	14 (34%)
Peripheral	66 (54%)	39 (48%)	27 (66%)
Discharge location				0.72
Home	15 (12%)	12 (15%)	3 (7%)
Acute rehabilitation	10 (8%)	7 (9%)	3 (7%)
Long-term facility	1 (1%)	1 (1%)	0 (0%)
Skilled nursing facility	6 (5%)	4 (5%)	2 (5%)
ECMO duration (hours)	94.8 (58.3-196.3)	110.8 (59.2-209.7)	93.9 (58.3-150.9)	0.58
Mortality	91 (74%)	58 (71%)	33 (80%)	0.24
Good neurologic outcome (mRS≤3)	16 (13%)	12 (15%)	4 (10%)	0.45

^*Ejection fraction categories according to American College of Cardiology/American Heart Association guidelines.

^**Variables were collected within the first 12 hours of ECMO initiation. For categorical variables which do not sum to 100%, the remainder consists of missing data. Abbreviations: SOFA score, arterial blood gas, and lactate measurements represent the worst value collected in the first 12 hours of ECMO. ABI: acute brain injury; SOFA: sequential organ failure assessment score; IABP: intra-aortic balloon pump; BP: blood pressure; BPV: blood pressure variability; SD: standard deviation; CV: coefficient of variation.

Overall, 58 (47%) patients had low pulse pressure (i.e., a median pulse pressure <20 mmHg in the first 12 hours of ECMO). Patients with ABI had a lower median systolic BP (92 vs. 96 mmHg, p=0.04) and lower median pulse pressure (20 vs. 32 mmHg, p=0.04) than those without (Figures 1B, S2). Diastolic BP (66 vs 64 mmHg, p=0.36) and MAP (74 vs. 74 mmHg, p=0.85) were similar in both groups. ABI patients were more likely to have low (<20 mmHg) pulse pressure (61% vs. 40%, p=0.03).

The best performing multivariable logistic regression model for predicting ABI included the variables: pulse pressure, cannulation strategy (central vs. peripheral), lactate on ECMO day 1 as a marker of disease severity, and left ventricle venting strategy. Only low pulse pressure (aOR=2.57, 95% CI=1.05-6.24, p=0.04) was independently associated with ABI (Table 2, Figure 2). When treated as a continuous variable in the same model, decreasing pulse pressure was associated with increased odds of having ABI (Figure 3A). Every 10-mmHg decrease in pulse pressure was associated with 31% increased adjusted odds of ABI (aOR=1.31, 95%CI=1.01-1.68, p=0.04) (Table S1). Furthermore, in the first 12 hours of ECMO, each additional 30 minutes spent with a pulse pressure <20 mmHg was associated with a 6% increased adjusted odds of ABI (aOR=1.06, 95%CI: 1.00-1.12, p=0.04) (Figure 3B). Exploratory analyses with different covariates are shown in the Supplement (Table S2-4). Pulse pressure was not associated with mortality or good functional neurologic outcome.

Table 2.

Multivariable logistic regression analysis of pulse pressure and acute brain injury.

Variable	aOR	95% CI	P value
Low pulse pressure (<20 mmHg)	2.57	1.05 – 6.24	0.04
Central cannulation	1.68	0.72 – 3.96	0.23
Lactate, ECMO day 1	1.05	0.98 – 1.12	0.19
LV venting strategy
None	Reference
IABP	0.82	0.30 – 2.24	0.69
Impella	0.31	0.05 – 2.02	0.22
Right superior pulmonary vein	0.25	0.03 – 1.79	0.17
LV apex	0.33	0.01 – 8.44	0.51

aOR: adjusted odds ratio; CI: confidence interval; LV: left ventricle.

Figure 2.

Forest plot for multivariable model of pulse pressure and acute brain injury (ABI), with the covariates of central cannulation, lactate on ECMO day 1, and left ventricle venting strategy. Dots represent adjusted odds ratios and brackets represent 95% confidence intervals. Low pulse pressure (<20 mmHg) was associated with ABI. LV: left ventricle; IABP: intra-aortic balloon pump.

Figure 3.

Multivariable logistic regression model with (A) pulse pressure treated as a continuous variable and (B) duration of low (<20 mmHg) pulse pressure, with the covariates of central cannulation, lactate on ECMO day 1, and left ventricle venting strategy. The probability of acute brain injury increases as (A) pulse pressure decreases and (B) duration of low pulse pressure increases.

In exploratory analysis separately analyzing ischemic and hemorrhagic stroke, we found that low pulse pressure trended towards significance (aOR=2.74, 95%CI: 0.86-8.75, p=0.089 for ischemic stroke; aOR=4.91, 95%CI: 0.77-31.4, p=0.093 for hemorrhagic stroke). However, we were limited by a low event rate for these individual diagnoses to make meaningful conclusions.

Subgroup Analysis

Out of 55 (45%) patients with an LV venting strategy, 35 (64%) had an IABP, 10 (8%) had an Impella, 7 (6%) and 2 (2%) had right superior pulmonary vein and left ventricle apex vents, respectively (Table 1). Those with LV venting had a significantly higher median pulse pressure than those without (38 vs. 16 mmHg, p<0.001). However, there were no statistically significant differences in ABI frequency between patients with an LV vent versus those without (25% vs. 40%, p=0.10). In a multivariable logistic regression model including only patients with LV venting, neither pulse pressure nor lactate on ECMO day 1 were associated with ABI (Table S5).

Out of 51 (42%) PCS patients, 19 (37%) experienced ABI. Median pulse pressure did not differ between patients with or without PCS (29 vs. 19 mmHg, p=0.12). Out of 44 (36%) non-PCS cardiogenic shock patients, 12 (27%) had ABI and there were no differences in median pulse pressure between those with or without non-PCS cardiogenic shock (28 vs. 20 mmHg, p=0.98). Pulse pressure, as a continuous or categorical variable, was not significantly associated with ABI in PCS or non-PCS cardiogenic stock patients (Table S5).

Systolic Blood Pressure and Blood Pressure Variability

Pulse pressure was non-linearly and strongly positively correlated with systolic BP (R²=0.84, RMSE=14) (Figure S3). As exploratory analysis, when MAP and systolic BP were added as covariates in separate multivariable ABI risk factor models, low pulse pressure was still significantly associated with ABI. Neither systolic BP nor MAP predicted ABI.

Compared to those without, patients with ABI had higher systolic (76.2 vs. 48.5, p=0.02) and diastolic (61.3 vs. 30.4, p=0.03) BPV as calculated by the coefficient of variation method (Table 1). BPV was very weakly positively correlated with pulse pressure (Figure S4). However, in a multivariable model with BPV as a covariate, low pulse pressure remained associated with ABI while neither systolic nor diastolic BPV (using either SD or CV calculation methods) were significantly associated. The distribution of systolic and diastolic BPV are shown in Figure S5.

Discussion

In our paper, we report an association of “low” pulse pressure after VA-ECMO cannulation with ABI by analyzing BP measurements taken every 15 minutes for the first 12 hours of ECMO support. Our cohort contained a high proportion of patients (33%) who sustained an ABI, reflective of the impact of standardized, multi-modal neurologic monitoring in early detection of ABIs for all patients on ECMO^4,5. Additionally, our cohort includes a high number of centrally cannulated patients (46%) who are at high risk of ABI²⁴. The median age (59 years), sex (63% male), and BMI (30 kg/m²) did not significantly differ between patients with ABI compared to those without. Among pre-ECMO comorbidities, only prior ICH was more frequent in patients with ABI (Table 1). After adjusting for lactate on ECMO day 1 (as a surrogate of baseline illness severity), LV venting strategy (since this can directly modulate pulse pressure), and cannulation strategy, low pulse pressure was an independent risk factor for ABI (Table 2). To our knowledge, this is the first report of the association of low pulse pressure with ABI in a VA-ECMO population.

In a histogram of pulse pressure (Figure 1) measurements across the first 12 hours of ECMO, two clusters were seen around pulse pressures of 15 mmHg and 45 mmHg. The first cluster is between 0 and 15 mmHg and is reflective of the subset of the population with severe cardiac dysfunction²⁵ who were found to have a higher rate of ABI. IABP was used in 28% of patients in our cohort, who had a median pulse pressure of 46 mmHg, which explains the bimodal distribution of pulse pressure because the counter-pulsation of IABP inflation and deflation can accentuate pulse pressure²⁶. We recognize that the use of IABP directly results in higher pulse pressure in contrast to Impella or direct LV venting, although those with IABP had the same rate of ABI compared to those without. After accounting for the hemodynamic influences of IABPs in multivariable analysis, we demonstrate that low pulse pressure remained significantly associated with ABI.

In the first 12 hours, patients with ABI had significantly lower median systolic blood pressure and lower pulse pressure, with a higher percentage of patients experiencing a pulse pressure <20 mmHg. Pulse pressure is a surrogate for underlying cardiac function and ejection, however, data on augmenting pulse pressure by ECMO blood flow or inotropes are weak²⁷. Pulse pressure can be augmented by inotropes, changing ECMO flow, or additional devices such as IABP, which alter arterial pressure and thereby afterload which has implications on native cardiac function and recovery²⁸. In animal models undergoing cardiac surgical operations, pulsatile extracorporeal circulation or cardiopulmonary bypass are associated with improved tissue blood flow and oxygenation when compared to non-pulsatile flow, which may predispose to both ABI and greater tissue damage after an ABI^29,30. Though the disadvantages of non-pulsatile flow are generally well-accepted, it remains the only option for VA-ECMO as existing pulsatile pumps are much less durable than continuous flow devices³¹.

Loss of pulsatile blood flow in the setting of mechanical circulatory support is associated with endothelial dysfunction, decreased local oxygen consumption, and disruption of cerebral autoregulation^32,33. Further, continuous flow mechanical support has been shown to induce von Willebrand factor deficiency³⁴. Taken together with intensive anticoagulation, VA-ECMO patients may have vasculature that is particularly prone to hemorrhage or thrombosis since they appear to respond improperly to blood gas fluctuations. These derangements at the microvascular level in the setting of disrupted autoregulation from loss of pulsatile flow are one possible explanation for the increased ABI rate. ECMO patients are subjected to a variety of supraphysiologic hemodynamic and blood gas conditions, such as hyperoxia and rapid carbon dioxide removal^6-8, which have been previously associated with ABI. The combination of these blood gas derangements in the setting of an already sensitive and disrupted endothelium provides a reasonable pathophysiological explanation of our findings. However, due to our limited sample size, we were unable to perform a multivariable analysis with blood gas and vital parameters such as PaO₂, change in PaCO₂, and hypothermia, which are important variables that influence ABI. Nonetheless, low pulse pressure was independently associated with a higher odds of ABI after adjusting pre-selected baseline severity of illness variables (i.e., lactate and SOFA score), operative approach, and LV venting strategy, reinforcing the important role pulse pressure plays as a dynamic marker of cardiovascular and neurologic injury.

BPV has been independently associated with adverse outcomes in the general population and those with atherothrombotic disease^9,22. Interestingly, both high¹⁸ and low¹⁹ BPV have been associated with poor prognosis. However, in our study, BPV was not associated with ABI. This may be due to several factors which differ in an ECMO population, including much worse underlying illness severity compared to previously studied populations, mechanical circulatory support leading to autoregulatory disruption as outlined above, the need for LV venting, and concomitant usage of multiple vasoactive drugs. Further, our cohort had lower blood pressure than previously studied, healthier populations, and BPV associations found previously likely do not extrapolate to the VA-ECMO population who are persistently at these low pressures.

Limitations

There are several limitations to this study. First, this is a retrospective single center analysis of a tertiary care center without a standardized approach towards pulse pressure, such as protocols for adjustments in ECMO flow or inotropes to achieve a set pulse pressure target. Due to our limited sample size, we were unable to control for many covariates which are known to be associated with neurologic outcomes in VA-ECMO, including blood gas derangements such as hyperoxia⁶ and large peri-cannulation carbon dioxide fluctuations^7,35, as well as temperature⁸. Additionally, although we attempted to control for baseline illness severity by using lactate and SOFA, these may not fully capture disease severity; likewise, we chose to control for LV venting strategy since they have a pronounced effect on pulse pressure, however, we did not control for ECMO flow rates or vasopressors and inotropes dosing (for example, through Vasoactive-Inotropic Scores) which may further augment pulsatility by affecting afterload. Our analysis was limited to the first 12 hours after cannulation with the suspicion that this was the time the brain was most vulnerable to acute change given a disruption of cerebral autoregulation; however, high quality data such as neuroimaging studies on critical time period for ABI occurrence has not been established. Finally, this was a retrospective study with the potential for immortal time bias, however, no ECMO patients in our study were withdrawn in the first 24 hours. Hence, the first 12 hours of pulse pressure and the time independent covariates in our model should not be affected. Larger cohorts, such as those utilizing the ELSO Registry, are better suited to Fine-Gray modeling to address death as a competing risk for ABI. In spite of these known limitations, our study is the first to examine the association of pulse pressure with ABI in an ECMO population.

Conclusions

In this single-center retrospective study, low pulse pressure, defined as < 20 mmHg through the first 12 hours of ECMO cannulation, was associated with ABI for patients on VA-ECMO support (Figure 4). Further research to establish specific pulse pressure targets and determine the ability to reduce incidence of brain injury by augmenting pulse pressure are warranted.

Figure 4.

Low pulse pressure (median <20 mmHg) in the first 12 hours after ECMO initiation is associated with a 2.57 increased adjusted odds of acute brain injury in patients supported with VA-ECMO.

Abbreviations

ABI

acute brain injury

aOR

adjusted odds ratio

BP

blood pressure

BPV

blood pressure variability

CI

confidence interval

CV

coefficient of variation

ECMO

extracorporeal membrane oxygenation

ELSO

Extracorporeal Life Support Organization

HIBI

hypoxic ischemic brain injury

IABP

intra-aortic balloon pump

ICH

intracranial hemorrhage

IQR

interquartile range

LV

left ventricle

mRS

modified Rankin Scale

PCS

post-cardiotomy shock

SD

standard deviation

SOFA

Sequential Organ Failure Assessment

VA-ECMO

venoarterial ECMO

HERALD (Hopkins Education, Research, and Advancement in Life support Devices) Investigators:

Kate Calligy, Patricia Brown, Diane Alejo, Scott Anderson, Matthew Acton, Hannah Rando, Henry Chang, Hannah Kerr