Recovery among post-arrest patients with mild-to-moderate cerebral edema

Abstract

Background:

Cerebral edema after cardiac arrest may be a modifiable cause of secondary brain injury. We aimed to identify processes of care associated with recovery in a cohort of patients with mild to moderate edema.

Methods:

We conducted a retrospective cohort study of adults resuscitated from out-of-hospital arrest (OHCA) at a single center from 2010–2018. We included those with cerebral edema ranging from mild to moderate (gray to white matter attenuation ratio (GWR) 1.2 to 1.3 on initial brain computerized tomography (CT). We used Pittsburgh Cardiac Arrest Category (PCAC) to adjust for illness severity and considered the following values in the first 24 hours of admission as additional predictors: GWR, lab values affecting serum osmolality (sodium, glucose, blood urea nitrogen (BUN)), total osmolality, change in osmolality from 0 to 24 hours, cardiac etiology of arrest, targeted temperature to 33°C (vs 36°C), time-weighted mean arterial pressure (MAP), partial pressures of arterial oxygen and carbon dioxide and select medications. Our primary outcome was discharge with Cerebral Performance Category 1 to 3. We used unadjusted and adjusted logistic regression for analysis.

Results:

We included 214 patients for whom CT was performed median 3.8 [IQR 2.4 – 5.2] hours after collapse. Median age was 57 [IQR 48–67] years, 82 (38%) were female, and 68 (32%) arrested from ventricular tachycardia or fibrillation. In adjusted models, modifiable processes of care were not associated with outcome.

Conclusions:

Illness severity, but not modifiable processes of care, were associated with recovery among post-arrest patients with mild-to-moderate cerebral edema.

Introduction

Early cerebral edema seen on computerized tomography (CT) imaging of the brain is an ominous sign after resuscitation from cardiac arrest.^1,2 Beyond marking injury severity, cerebral edema may contribute to preventable secondary brain injury by compromising capillary blood flow, increasing capillary-tissue oxygen gradients, or raising intracranial pressure.^3,4 These effects may decrease cerebral blood flow and oxygen delivery.^3,4 Edema severity can be approximated by the ratio of grey matter attenuation to white matter attenuation (GWR) on CT imaging. Previous studies have shown that severe edema (GWR ≤ 1.1) likely reflects irrecoverable brain injury.^2,5 These studies found variable recovery among patients with GWR 1.2–1.3.^2,5

Whether mild to moderate cerebral edema is a modifiable contributor to secondary brain injury is unknown. We analyzed a large cohort of subjects with potentially recoverable early cerebral edema, described their baseline clinical phenotypes, and sought to identify modifiable factors associated with survival without severe neurological impariment. Specifically, we hypothesized that higher sedation dose, temperature management to 33°C and rising serum osmolality would be associated with improved outcome.

Methods

Setting and Population

The University of Pittsburgh Human Research Protection Office approved this study. We performed a retrospective observational cohort study including consecutive patients admitted to a single academic medical center in Western Pennsylvania after resuscitation from out-of-hospital cardiac arrest (OHCA) with GWR 1.2–1.3 on initial brain CT. In a secondary analysis, we included patients with GWR 1.2–1.4. We excluded patients that were awake (following verbal commands) or died within 6 hours of return of spontaneous circulation (ROSC), those that did not have computerized tomographic (CT) imaging of the brain within 24 hours of ROSC, those <18 years of age, and those that arrested due to trauma or a primary neurological event such as intracranial hemorrhage. It is our clinical standard of care to obtain CT imaging of the brain on all comatose OHCA patients. We do not obtain imaging for patients with prior wishes are prohibitive of critical care, or in rare circumstances where overt clinical instability prevents safe imaging (e.g. need for emergent mechanical circulatory support). We maintain a prospective registry including all patients treated by our Post-Cardiac Arrest Service (PCAS) that we used to identify patients. We have previously described the role of this service and our standard clinical practices in detail.^6–8

Predictors and outcomes

For each patient, we calculated GWR by dividing the density of the caudate nucleus (grey matter) in Hounsfield units by the density of the internal capsule.² We substituted the density of the suboccipital grey matter if the internal capsule was poorly visualized.² We recorded GWR to two decimal places. We used GWR to determine eligibility for inclusion, as above, but also included it as a predictor because we observed a strong association of outcome even between 1.2 and 1.3. We quantified post-arrest illness severity using the Pittsburgh Post-Cardiac Arrest Category (PCAC), a validated measure of post-arrest illness severity.⁸ We classified arrest etiology as cardiac vs other, as we have previously described,⁹ and included other standard Utstein-style predictors: witnessed collapse, CPR duration, and number of epinephrine doses administered. We selected additional covariates based on biological plausibility. These included target temperature management (TTM) to 33°C compared to 36°C, because deeper hypothermia might reduce both edema and cerebral metabolic rate of oxygen.¹⁰ Similarly, we considered mean arterial pressure (MAP) as a predictor. It is our routine practice to place an arterial catheter for invasive blood pressure monitoring. Intensive care unit nurses record MAP and other vital signs at least hourly. We performed an electronic extraction of all vital signs and lab values from the electronic medical record and calculated a time-weighted mean MAP over the first 24 hours after admission by carrying the last documented MAP value forward until a new value was documented. We typically sample arterial blood gases every 4–6 hours in the first day after cardiac arrest, and calculated time-weighted averaged in partial pressure of arterial oxygen (PaO₂) and carbon dioxide (PaCO₂). Since change in serum osmolality may alter cerebral edema, we further considered osmotically active serum solutes as predictors. These included serum sodium, albumin and glucose levels, as well as estimated total serum osmolality, which we calculated as 2*[Sodium] + [Glucose]/18 + [Blood urea nitrogen]/2.8, over the first 24 hours. We also considered the change in these values from presentation to 24 hours, since the tonicity of extravascular brain water is slow to equilibrate with serum and rapidly rising serum osmolality may reverse cerebral edema. Finally, we considered use of propofol for sedation, mannitol, hypertonic saline, antiplatelet therapy, and anticoagulants. Mannitol and hypertonic saline were administered based on suspicion of increased intracranial pressure from an attending physician. We treated all medications as a binary predictors, without consideration of dose.

Our primary outcome was survival without severe neurological impairment, which we defined as Cerebral Performance Category (CPC) of 1–3 at discharge. A PCAS attending physician assessed each patient daily in the absence of sedation (unless sedation interruption was contraindicated, for example in the context of refractory hypoxemia requiring neuromuscular blockade) until death or discharge.

Statistical Analysis

We used descriptive statistics to summarize population characteristics. We used unadjusted and adjusted logistic regression to test the association of each predictor with our outcome. We included variables in the adjusted models with significant univariable associations with outcome (P < 0.05). As a secondary analysis we used inverse probability weighting to estimate average treatment effects for mean MAP, PaCO₂, PaO₂, TTM to 33°C and osmolar therapy (combined mannitol and hypertonic saline) using other available covariates to predict the probability of exposure or treatment. Finally, we repeated our analyses restricting the cohort to those with CT imaging obtained within 6 hours of collapse. We used STATA version 15.1 (StataCorp, College Station, TX) for all analyses.

Results

We identified 1,500 out-of-hospital cardiac arrest during our study period of which 214 met inclusion and exclusion criteria (Figure 1). Median age was 57 [interquartile range (IQR) 48–67] years, 82 (38%) were female, and the most common presenting rhythm was ventricular tachycardia or fibrillation (32%) (Table 1). The large majority of CT scans (88%) were obtained within 6 hours of collapse. CPC 1–3 was seen in 43 (21%) patients, with the median length of stay for survivors 15 [IQR 9–22] days.

Figure 1:

Consort diagram showing patient inclusion in current study

Table 1:

Baseline population characteristics

Characteristic	Median [IQR], N (%)
Age, years	57 [48–67]
Sex, female, N (%)	82 (38)
Grey White Ratio	1.27 [1.25–1.29]
Hours from collapse to CT	3.8 [2.4 – 5.2]
Pittsburgh Cardiac Arrest Category, N (%)
2	43 (20)
3	17 (8)
4	140 (65)
Unknown	14 (7)
Arrest Rhythm
VT/VF	68 (32)
PEA	65 (30)
Asystole	65 (30)
Unknown	16 (8)
MAP, mmHg	89 [82 – 95]
PaCO2, mmHg	42 [37 – 47]
PaO2, mmHg	163 [122 – 209]
Etiology
Cardiac	46 (22)
Non-Cardiac	168 (78)
Witnessed Arrest	139 (65)
Duration of CPR (min)	18 [10–29]
Epinephrine doses administered	3 [1 – 4]
Survived to hospital discharge	51 (24)
Cerebral Performance Category at Discharge
1	9 (18)
2	6 (12)
3	28 (55)
4	8 (15)
Modified Ranking Scale at Discharge
0	3 (6)
1	5 (10)
2	5 (10)
3	9 (18)
4	18 (35)
5	11 (21)
Length of Stay
Survivors	15 [9–22]
Nonsurvivors	3 [2–5]

VT – Ventricular Tachycardia, VF – Ventricular Fibrillation, PEA – Pulseless Electrical Activity

In unadjusted models, multiple factors were associated with our outcome (Table 2). Antiplatelet therapy was collinear with cardiac etiology of arrest, so it was not included in adjusted models. In adjusted models, PCAC 4 vs 2 (adjusted OR 0.11, 95% CI (0.03–0.45) was negatively associated with outcome, but no predictors were positively associated with outcome. Modifiable processes of care including TTM, MAP, PaO₂, PaCO₂, hypertonic medication administration and propofol use were not independently associated with our outcome in our adjusted analysis. When we broadened inclusion criteria to patients with GWR 1.2–1.4, 513 met inclusion criteria, of whom 132 (26%) had a discharge CPC 1–3. Adjusted models were substantively similar (Supplemental Table 1). Results from our secondary analysis using inverse probability weighting to estimate average treatment effects were similar, without a significant association between any treatment and outcome, as were results when we restricted analysis to patients with CT imaging within 6 hours of collapse.

Table 2:

Adjusted and Unadjusted logistic regressions predicting Survival without devastating neurological impairment GWR 1.2–1.3

	Unadjusted		Adjusted
Predictor	OR (95% CI)	P	OR (95% CI)	P
Age, per 5 years	0.98 (0.88 – 1.08)	0.666	-	-
Female sex	0.83 (0.41 – 1.67)	0.605	-	-
Pittsburgh Cardiac Arrest Category
2	Ref		Ref
3	0.67 (0.21 – 2.08)	0.487	-	-
4	0.05 (0.02 – 0.13)	<0.001*	0.11 (0.03 – 0.45)	0.002*
Target temperature
36C	Ref		Ref
33C	0.27 (0.13 – 0.57)	0.001*	0.30 (0.09 – 1.06)	0.061
GWR, per 0.01	1.31 (1.12 – 1.52)	0.001*	1.26 (0.98 – 1.61)	0.067
Day 1 laboratory and physiological values
Serum sodium, per 5 mEq/L	0.88 (0.58 – 1.34)	0.554	-	-
Albumin, per 1 g/dL	1.64 (0.90 – 2.98)	0.103	-	-
Glucose, per 25 mg/dL	0.82 (0.71 – 0.96)	0.013*	0.77 (0.55 – 1.08)	0.124
Blood urea nitrogen, per 5 mg/dL	0.79 (0.67 – 0.92)	0.003*	0.83 (0.63 – 1.08)	0.166
Osmolality, per 10 mOsm/kg	0.54 (0.36 – 0.80)	0.002*	1.32 (0.48 – 3.61)	0.592
Mean arterial pressure, per 10 mmHg	1.20 (0.92 – 1.58)	0.185	-	-
PaCO2, per 5 mmHg	1.04 (0.86 – 1.25)	0.683	-	-
PaO2, per 5 mmHg	0.99 (0.97 – 1.03)	0.868	-	-
Medication administration
Anticoagulant	1.34 (0.61 – 2.91)	0.467	-	-
Antiplatelet	2.30 (1.12 – 4.74)	0.023	-	-
Hypertonic saline	1	-	-	-
Mannitol	1	-	-	-
Propofol	1.53 (0.72 – 3.26)	0.267	-	-
Change in osmolality from day 1 to 2, per 10 mOsm/kg	0.53 (0.31 – 0.88)	0.015*	0.33 (0.06 – 1.85)	0.206
Change in sodium from day 1 to 2, per 5 mEq/L	0.46 (0.25 – 0.82)	0.009*	1.51 (0.23 – 9.84)	0.664
Cardiac etiology	3.68 (1.77 – 7.62)	<0.001*	2.84 (0.84 – 9.60)	0.094
Epinephrine doses administered	0.65 (0.53 – 0.79)	<0.001*	0.79 (0.57 – 1.09)	0.146
CPR duration	0.95 (0.92 – 0.98)	0.001*	0.99 (0.96 – 1.03)	0.666
Witnessed arrest	1.74 (0.82 – 3.69)	0.149	-	-

GWR – Grey White Ratio

Discussion

We did not identify modifiable processes of care associated with recovery among patients with mild-to-moderate cerebral edema. Both cytotoxic and vasogenic components contribute to edema formation in this population.^3,11 Disruption of the blood brain barrier (BBB) from oxidative stress or inflammation, for example, results in extravasation of large osmotically active molecules and fluid typically excluded from the extracellular space, although some data suggest that ionic shifts and active transport via aquaporins play a larger role than BBB disruption in this population.^3,12 Concurrently, anoxia causes energetic failure resulting accumulation of intracellular sodium and water. In severe hypoxic-ischemic injury, cellular necrosis disrupts all homeostasis and presents radiographically as severe edema.

Cerebral edema is a potentially modifiable cause of secondary brain injury after cardiac arrest.¹³ Global mass effect from severe edema can increase intracranial pressure and reduce cerebral blood flow.^3,4 In more mild cases, localized edema of perivascular astrocyte end-feet can impair microcirculatory flow and create a diffusion barrier that contributes to brain tissue hypoxia.⁴ Target temperature management (TTM) is widely used in post arrest populations and may reduce cerebral edema through multiple mechanisms.^14,15 We did not see an association between TTM and outcomes in our adjusted models.

Increasing serum osmolality through administration of hypertonic medications reduces cerebral edema in non-arrest populations by drawing water out of brain tissue.¹⁶ Animal models of cardiac arrest have shown anti-edema effects of hyperosmolar therapy in both mice and swine.^17,18 However, with damage to the BBB, as seen in vasogenic edema, hyperosmolar therapy could exacerbate cerebral edema as hypertonic solutes extravasate into the interstitium.¹⁹ Since cerebral edema after cardiac arrest is multifactorial, damage to the BBB could potentially help explain why we did not see any survival benefit of hyperosmolar therapy in our current study. Selection bias, whereby sicker patients either received hypertonic agents or had natural increases in serum osmolality over time, from metabolic disarray or end organ failure, may also be confounders.

Our study has several limitations. The single center, retrospective design limits generalizability to other institutions. In selecting our cohort, we classified edema based on CT for pragmatic reasons because of its wide availability and rapid, logistically simple acquisition in unstable patients with ongoing resuscitation. However, compared to magnetic resonance imaging (MRI), CT is insensitive to mild or localized edema. Thus, we cannot comment on the potential efficacy of the predictors we evaluated on more subtle edema inapparent on CT. We included patients who received a head CT within the first 24 hours of their arrest, but did not consider time from collapse to imaging within that range. Brain CT within the first 2 hours of ROSC may not predict outcome.²⁰ Treating radiographic measures of edema as time-invariant oversimplifies a complex physiological process and may have introduced unmeasured heterogeneity to our cohort. Similarly, MAP, oxygenation and ventilation are complex continuous physiological exposures that are only sampled intermittently. Our consideration of average daily values may overlook clinically important phenomena (e.g. an hour of severe hypoxia or hypotension may be harmful despite a normal average value). Because clinicians work to keep these values within relatively normal target ranges, the observed variability in these parameters (Table 1) may have been insufficient to detect associations with outcomes. Complex causal relationships between predictors and outcome or heterogeneity between patients may also obscure a protective. For example, higher MAP may improve cerebral blood flow and outcome in one patient, but be caused by impending herniation in another. Operationalizing exposure to these physiological parameters in statistical models is an area of ongoing investigation.

When selecting laboratory values as candidate predictors of recovery, we chose those that would be commonly drawn and readily available to providers. Thus, our final variable list is not comprehensive. One value which we did not include was neuron specific enolase (NSE), which other studies have used in their models predicting recovery of patients post arrest, since it is not routinely checked at our center.²¹ Because relatively few favorable outcomes were observed in this cohort, there is a risk of model overfitting. This and multicollinearity may explain both the observed instability in parameter estimates between unadjusted and adjusted models and lack of significant independent predictors of outcome. Inverse probability weighted treatment effect estimation only incompletely overcomes this limitation.

Overall our study supports that cerebral edema after cardiac arrest is an ominous sign. Current therapies as delivered in routine clinical care appear inadequate. More research is needed to elucidate the underlying mechanisms of edema formation, developing more sensitive approaches to detection that are feasible to use in unstable patients and identify individualized therapeutic targets.