Cerebral Edema in Traumatic Brain Injury: a Historical Framework for Current Therapy

Abstract

Purpose of review

The purposes of this narrative review are to (1) summarize a contemporary view of cerebral edema pathophysiology, (2) present a synopsis of current management strategies in the context of their historical roots (many of which date back multiple centuries), and (3) discuss contributions of key molecular pathways to overlapping edema endophenotypes. This may facilitate identification of important therapeutic targets.

Recent findings

Cerebral edema and resultant intracranial hypertension are major contributors to morbidity and mortality following traumatic brain injury. Although Starling forces are physical drivers of edema based on differences in intravascular vs extracellular hydrostatic and oncotic pressures, the molecular pathophysiology underlying cerebral edema is complex and remains incompletely understood. Current management protocols are guided by intracranial pressure measurements, an imperfect proxy for cerebral edema. These include decompressive craniectomy, external ventricular drainage, hyperosmolar therapy, hypothermia, and sedation. Results of contemporary clinical trials assessing these treatments are summarized, with an emphasis on the gap between intermediate measures of edema and meaningful clinical outcomes. This is followed by a brief statement summarizing the most recent guidelines from the Brain Trauma Foundation (4th edition). While many molecular mechanisms and networks contributing to cerebral edema after TBI are still being elucidated, we highlight some promising molecular mechanism-based targets based on recent research including SUR1-TRPM4, NKCC1, AQP4, and AVP1.

Summary

This review outlines the origins of our understanding of cerebral edema, chronicles the history behind many current treatment approaches, and discusses promising molecular mechanism-based targeted treatments.

Introduction

Cerebral edema (defined as an increase in the water content of central nervous system [CNS] cells or interstitium) and accompanying intracranial hypertension have long been recognized as important contributors to morbidity and mortality after traumatic brain injury (TBI) [1–13]. Mass effect from cerebral edema may cause fatal herniation, increase intracranial pressure (ICP), decrease cerebral perfusion pressure (CPP), and compromise cerebral oxygenation [2]. A significant proportion of clinical and financial resources following TBI are spent on the identification and management of cerebral edema, intracranial hypertension, and their consequences [1, 4, 14].

The pathophysiology underlying cerebral edema remains incompletely understood. This precludes optimal clinical management regarding whether, when, and how to treat it. Like neuroinflammation, some degree of edema, or molecular aspects of its development, may promote repair and regeneration, but cause injury in extreme forms [15]. The decision to treat cerebral edema, and measurement of treatment efficacy, is frequently guided by ICP. However, ICP is an imperfect proxy for edema and does not reflect underlying mechanisms. Although clinical studies of different therapies targeting brain edema have demonstrated improvements in ICP [16–18], or radiographic and serum biomarkers associated with swelling [19, 20], few demonstrate improved functional outcomes. Noted functional benefits have largely represented patient shifts from deceased to severely disabled categories [21••, 22].

Here, we use a historical framework (Fig. 1) to provide an understanding of the evolution of current treatment of cerebral edema after TBI, the data supporting these practices, and molecular mechanism-based therapies being developed. We also briefly discuss mechanisms underlying cerebral edema and the use of ICP as a proxy, this is done in the context of addressing current practices, limitations, and future treatment strategies. Although further development of serum and imaging biomarkers for cerebral edema is key to a precision medicine approach to treatment, this is beyond the scope of this review.

Fig. 1.

A historical timeline of strategies to treat cerebral edema and intracranial hypertension after traumatic brain injury. This schematic highlights key discoveries in the evolution of treatment strategies for cerebral edema and intracranial hypertension including decompressive craniectomy (pink text boxes), external ventricular drainage (yellow text boxes), hyperosmolar therapy (blue text boxes), sedation (orange text boxes), and hypothermia (green text boxes). Recent landmark randomized controlled trials for each of these categories are also marked. Time scale is shown on the gray line (not drawn to scale).

Pathophysiology of cerebral edema—implications for therapy and clinical decision making

Starling forces

In 1783, the Scottish surgeon Alexander Monro first described the nature of fixed intracranial volume:

For, as the substance of the brain, like that of the solids of our body, is nearly incompressible, the quantity of blood within the head must be the same, or very nearly the same, at all times, whether in health or disease, in life or after death, those cases only excepted in which water or other matter is effused or secreted from the blood vessels; for in these, a quantity of blood, equal in bulk to the effused matter, will be pressed out of the cranium.[23]

The Monro-Kellie hypothesis has had several iterations and now expresses the obligate relationship between volumes of brain, cerebrospinal fluid (CSF), and blood within the fixed cranial vault. Pertaining to cerebral edema, an increase in brain parenchyma volume must be matched by a corresponding decrease in the volume of blood and CSF. This concept underlies current management strategies: sedatives and hypothermia decrease cerebral blood flow (CBF) and thus cerebral blood volume (CBV); hypertonic and hyperosmotic solutions decrease both brain parenchymal volume and CBV; external ventricular drainage reduces CSF volume; and decompressive surgery increases the naturally fixed cranial volume.

The physical drivers of cerebral edema, like other forms of systemic edema, are related to the modified Starling equation (Box-1) [24]. Tight junctions render the blood-brain barrier (BBB) nearly impermeable under physiological conditions, though it is often disrupted in TBI [25–27]. This may contribute to multiple edema subtypes: vasogenic, cytotoxic, and ionic, as discussed below [28, 29]. Another consequence of this normally low conductivity is the significance of aquaporins (AQP). Because most systemic vascular beds are porous, AQP contributes minimally to edema and fluid flux in most tissues [30]. However, due to the near impermeability of the BBB, AQP4 likely has a greater role in both the pathogenesis and resolution of edema following TBI, which may have important therapeutic implications [31–33].

Categories of cerebral edema

Cerebral edema is classically categorized into three broad subtypes: oncotic, vasogenic, and cytotoxic [4, 26]. While the overlap and interdependence of edema subtypes (and progression to contusion expansion) is increasingly recognized, the relative contributions of different molecular mechanisms in various disease states may result in one subtype predominating over others. For example, CNS neoplasm-related edema is considered primarily vasogenic, a consequence of tumor-driven activation of angiogenic pathways, or the body’s inflammatory response [34, 35]. TBI results in both vasogenic and cytotoxic edema, with variation depending on injury mechanism, time, contributory molecular processes, intracranial injury patterns, and individual differences [3, 28, 36–42]. These differences inform approaches to targeted therapy development and management strategies.

Oncotic edema

Oncotic edema relates primarily to the second half of the Starling equation: fluid shifts result from vascular hypoosmolarity or interstitial hyperosmolarity. As described by Katayama et al., interstitial hyperosmolarity after TBI is common, with disruption of ion homeostasis across endothelial cell membranes, cell death due to various mechanisms, and contusion necrosis [26, 43, 44]. Serum hypoosmolarity after TBI may also occur (e.g., hyponatremia from the syndrome of inappropriate antidiuretic hormone [45] or the less common cerebral salt wasting [46]). Because the modified Starling equation depends on the osmolarity of the subglycocalyx space, rather than only the osmolarity of the interstitium directly, CNS osmotic edema is also dependent to some extent on the disruption of the separation between these compartments. The BBB is often disrupted early after TBI, particularly in contused brain regions, but can be re-established [43]. This can contribute importantly to contusion expansion as contusion necrosis increases local tissue osmolar load.

Vasogenic edema

Vasogenic edema relates to BBB disruption: fluid extravasates from the vascular compartment to the interstitial space with an increase in the endothelial membrane hydraulic conductivity [47]. BBB disruption may result from damage to the vascular endothelium or astrocytic foot processes, or from pathologic signaling pathways upregulating vasoactive mediators such as tissue necrosis factor (TNFα), vascular endothelial growth factor (VEGF), vasopressin, matrix metalloproteinases (MMPs), and sulfonylurea receptor 1-transient receptor potential cation channel subfamily member (SUR1-TRPM4) [4, 48–54]. Although the molecular overlap between neuroinflammation and cerebral edema related to vascular permeability is beyond the scope of this review, a potential teleological role for post traumatic hyperemia and vasogenic edema may be to facilitate clearance of debris and delivery of beneficial complement components, cytokines, and other neuroinflammatory mediators. These promote trophic factors that support neuroregeneration, integration, and repair.

Cellular/cytotoxic edema

The role of cellular/cytotoxic edema after TBI is complex and evolving. It is both a protective homeostatic process and has detrimental consequences when it results from ATP depletion [55]. After TBI, glutamate release and hyperglycolysis often occur in the absence of ischemia/ATP depletion [56]. Astrocytes take up toxic-metabolic products of these processes, which may limit their impact on the neuronal environment, thereby protecting neurovascular elements and promoting recovery [55–57]. However, this also results in significant astrocyte swelling due to increased intracellular solute loading [55]. This may compromise other cellular functions and has unintended downstream consequences.

In addition to homeostatic mechanisms of astrocytic cellular swelling, cytotoxic (cellular) edema can result from failure of ion homeostasis across cell membranes [43, 58, 59]. ATP depletion results in failure of multiple ion pumps, chief among them being the Na/K ATPase [60], with multiple downstream consequences including voltage-gated channels opening from membrane depolarization, and impairment of sodium-dependent secondary active transport [61]. ATP depletion further results in increased ion flux into the cells through channels such as SUR1/TRPM4 and NKCC1, disrupting tightly regulated ion gradients [62, 63]. This process occurs across CNS cell types including neurons, vascular endothelium, and astrocytes. Dysregulated ion gradients cause water to shift into the intracellular space from the interstitium, resulting in cellular swelling and disruption of intracellular processes [28, 29, 43, 59]. Unchecked, this may result in cell death. Fluid redistribution from the interstitium to the cellular compartment would cause cellular edema but would not increase total brain volume/water. However, oncotic cell death, particularly in endothelial cells or astrocytes, increases BBB disruption, thus facilitating an increase in total brain water via osmotic and vasogenic mechanisms.

Cerebral edema and ICP

Although some forms of cerebral edema can be identified radiographically, clinical treatment decisions are frequently made based on ICP [64]. While all cerebral edema may not warrant intervention, equating cerebral edema with increased ICP has important limitations. Some patients may not develop intracranial hypertension from cerebral edema due to age, atrophy, or intracranial compliance. Because the intracranial volume-pressure curve is non-linear—the curve remains nearly flat up to a critical point, at which point it increases exponentially [2]—a patient with significant edema burden may rapidly develop life-threatening intracranial hypertension as they approach the inflection point on the curve. In selected cases, intracranial hypertension after TBI may result from an increase in cerebral blood volume rather than edema (although edema has been shown to constitute the predominant cause) [65]. Conversely, edema may accumulate without intracranial hypertension. The extent to which cerebral edema generation without intracranial hypertension is harmful has just begun to be explored [66]. This likely depends on the underlying pathways and their effects on neuronal or glial death. Recent data suggest that certain ICP trajectories increase odds of unfavorable outcomes in the absence of intracranial hypertension [67]. Analysis of various characteristics of ICP waveforms can also inform on cerebral compliance [68–71]. However, methods to best incorporate this information into clinical practice remain unclear. Although not discussed here, there is also a body of literature challenging the notion of ICP-based measurement/treatment improving clinical outcomes after severe TBI [19, 72–74].

Treatment of intracranial hypertension and cerebral edema

Decompressive craniectomy

Evidence of craniotomy and trephination goes back 8000–10,000 years in what is now Peru and has been found throughout the ancient world including Egypt, and Central and South America [75]. The first written document describing therapeutic trephination is Hippocrates’ On Injuries of the Head, from approximately 400 BCE [76, 77]. While not explicitly writing that trephination was performed to relieve clot or pressure, he stated that “the more the bones are contused and depressed and comminuted, the less they require trephining” [77]. Hundreds of years later in the nineteenth century, with the advent of antiseptic techniques and general anesthesia, that the procedure regained popularity for management of primary lesions and intracranial hypertension [78].

Contemporary data on the use of decompressive craniectomy to treat intracranial hypertension after TBI derive from multiple observational studies and three randomized controlled trials (RCTs, Table 1) [21••, 79–90]. The results have been mixed. While decompressive craniectomy effectively reduces ICP, and possibly length of ICU stay, effects on functional outcome are less clear. The Brain Trauma Foundation guidelines provide a level IIA recommendation that decompressive hemicraniectomy not be used for improving outcome, while acknowledging that the procedure may reduce ICP and ICU stay [64]. These guidelines were published prior to the RESCUEicp trial (Table 1). A subsequent consensus statement from the International Consensus Meeting on the Role of Decompressive Craniectomy in the Management of TBI (2019) acknowledged that decompressive surgery may effectively reduce ICP and mortality [91]. They also emphasized that this may only be beneficial for a subset of patients, and that candidates should be carefully selected after considering the entire clinical picture. Recent preclinical studies have raised concern that decompressive craniectomy reduces glymphatic flux, thus decreasing clearance of toxic metabolites from the injured brain, resulting in gliosis and neuroinflammation [92].

Table 1.

Randomized controlled trials of decompressive craniectomy after TBI

Authors (year)	Study title	Sample size	Primary outcome	ICP result	Study limitations
Taylor et al.2001 [1]	A randomized trial ofvery early decompressivecraniectomy in children with traumatic brain injury and sustained intracranial hypertension	n = 27	Favorable 6-month GOS in 14% of medically managed patients, 54% of craniectomy patients. Not statistically significant	Average 48-h ICP decrease in surgery arm of 8.98 (SD 6.60) and a decrease of 3.69 (SD 7.14) in the medical arm. Not statistically significant	Small sample size Authors report changes in standard of care management during the study
Cooper et al. 2011 [2]	Decompressive craniectomy in diffuse traumatic brain injury (DECRA)	n = 155	Odds ratio of unfavorable 6-month GOS-E in craniectomy group was 1.84 (1.05–3.24 95% CI)	Patients undergoing surgery required fewer interventions for ICP control (p < 0.05) Mean post-randomization ICP was lower in the surgery group (14.4 mmHg vs 19.1 mmHg, p < 0.05) Proportion of end-hourly ICP readings over 20 was lower in surgery group (p < 0.05)	All patients received large bilateral operations without surgeon input 18% of patients in medical arm crossed-over Admission pupil reactivity was decreased in surgery arm Outcome difference no longer significant after post hoc adjustment for pupil reactivity
Hutchinson et al. 2016 [3]	Trial of Decompressive Craniectomy for Traumatic Intracranial Hypertension (RescueICP)	n = 389	6-month GOS-E was higher in in surgery group than control group (p < 0.05)	Patients in the surgery arm had lower median ICP post-randomization (14.5 mmHg vs 17.1 mmHg respectively, p < 0.05) Patients in the surgery arm had fewer hours with ICP > 25 after randomization (5.0 vs 17.0, p < 0.05) Proportions of end-hourly ICPs above both 20 and 25 were significantly lower in the surgery arm (p < 0.05)	Difference in GOS-E scores at 6 months was driven primarily by movement within the broader category of poor GOS-E (prespecified as lower severe disability or worse) Proportion of patients achieving good outcome at 6 months did not achieve statistical significance

Favorable GOS: GOS > 3; Favorable GOS-E: GOS-E > 4; landmark studies in italics

GOS Glasgow outcome scale, GOS-E Glasgow outcome scale-extended, ICP intracranial pressure

External ventricular drains

The first documented external ventricular drain (EVD) procedure was performed in 1744 by Claude-Nicolas Le Cat with a ventricular puncture and wick in place to drain CSF [93]. In 1960, Nils Lundberg systematically described ICP monitoring techniques and confirmed feasibility in 30 TBI patients [94, 95]. Lundberg also characterized two types of ICP fluctuations which still hold clinical and prognostic significance today: Lundberg “A” waves—sustained increases in ICP to levels > 50 mmHg for 5–20 min before rapid resolution, and Lundberg “B” waves—0.5–2 Hz rhythmic oscillations to ICPs 20–30 mmHg above baseline but without sustained ICP increases [95]. These patterns represent significantly reduced intracranial compliance. Nonetheless, EVD use remained uncommon in TBI until 1982 when Narayan et al. demonstrated a link between intracranial hypertension and outcome following TBI and recommended prompt treatment [96]. EVD use for ICP monitoring remains controversial even today. Multiple cohort studies report an association between ICP monitoring and decreased mortality [97–99]. However, the Benchmark Evidence from South American Trials: Treatment of Intracranial Pressure (BEST:TRIP) trial, the only RCT to compare invasive monitoring to clinical exam and imaging-guided care, found no significant difference in either mortality or functional outcome between groups [19]. Some reported limitations of this study include an equal weighting of mortality in its 21-point composite outcome, insufficient power to assess mortality, access to pre-resuscitation care, and difficulty generalizing to the USA where management protocols and TBI care differ.

EVDs are used after TBI for both ICP monitoring and therapeutic CSF drainage; impact of the latter has not been studied via RCT in TBI. Retrospective studies suggest that continuous CSF drainage is associated with decreased ICP burden and improved cerebral perfusion pressure [100–104]. However, when comparing pressure monitoring via EVD to intraparenchymal catheters (IPC), a retrospective cohort study of data from the Citicholine Brain Injury Treatment Trial found that EVDs were associated with lower GOSE scores, higher mortality, and lower performance across neuropsychological parameters [105]. A 2015 meta-analysis found no difference between EVDs and IPCs with regard to mortality or functional outcome, but reported higher complication rates with EVD use [106].

The Brain Trauma Foundation guidelines provide a level III recommendation that an EVD levelled to the midbrain be considered in patients with a GCS score < 6 within 12 h of injury to lower ICP, and recommend that the EVD be allowed to drain continuously, rather than intermittently [64]. Additional studies comparing outcomes for ICP-directed care based on EVD vs parenchymal monitors are needed.

Hyperosmolar therapy

Although Monro and Kellie suggested that the volume of the brain was immutable, the foundational work of Weed and McKibben from 1919 [23, 107–109] demonstrated that intravenous injections of hypotonic fluids, such as water, cause marked increases in both brain volume and ICP, while injections of hypertonic fluids, such as hypertonic saline or hypertonic sodium bicarbonate solutions reduce these same parameters [107, 108]. After the advent of ICP monitoring in 1927, several hypertonic solutions were tested for controlling ICP: 50% glucose, 50% sucrose, 25% sodium chloride, 50% magnesium sulfate, hyperoncotic albumin, and hyperoncotic plasma [110]. However, these often resulted in rebound intracranial hypertension. Thus, hyperosmotic agents fell out of favor until 1954 with the discovery of hypertonic urea [111–113]. Mannitol was first administered in 1960 [114] and rapidly overtook urea as the osmotic agent of choice for ICP control because it was readily soluble and easy to store.

In 1988, Worthley and colleagues reported on hypertonic saline for ICP control in patients with acute kidney injury following repeated mannitol administration [115]. Thereafter, hypertonic saline was increasingly used in patients with intracranial hypertension refractory to mannitol, or with coexisting hypovolemic shock given its volume expansive properties [116–119]. The ability to continuously infuse hypertonic saline contributed significantly to its resurgence as a treatment option for raised ICP, particularly in children [119, 120].

Contemporary studies have demonstrated that both mannitol and hypertonic saline effectively reduce ICP, and many have compared the two, as summarized (Table 2). Bolus hypertonic saline therapy may control ICP more effectively than mannitol, though the clinical significance is questionable (with absolute differences of ~ 2 mmHg) [121]. None of these studies demonstrates a difference in mortality or functional outcome [17, 122–131]. No RCTs compare continuous hypertonic saline infusion with bolus hyperosmotic therapy. Cerebral adaptation to a prolonged hyperosmolar state may reduce the oncotic pressure gradient and limit efficacy of acute ICP reduction from continuous infusions after the first 48 h [4, 50, 132]. The Brain Trauma Foundation guidelines for adults do not offer a recommendation on the use of hyperosmotic agents. They acknowledge the key role that these agents play in clinical practice, but that the current state of evidence does not allow for a formal recommendation [64]. However, the Brain Trauma Foundation guidelines for severe pediatric TBI recommend bolus administration of hypertonic saline based on the aforementioned studies, as well as more recent studies comparing bolus hypertonic saline to other therapies targeting intracranial hypertension, such as narcotics or barbiturates [133, 134].

Table 2.

Randomized controlled trials of hypertonic saline and mannitol after TBI

Authors (year)	Study title	Study size	Primary outcome result	ICP result	Study limitations
Fisher et a1.1992 [4]	Hypertonic Saline Lowers Raised Intracranial Pressure in Children After Head Trauma	n = 18	3% HTS was associated with ↓ ICP over 2 h, while 0.9% saline was not (p < 0.05)	Average ICP during infusion of 3% HTS was lower than it was during infusion of 0.9% saline (18.7 vs 14.0 mmHg respectively, p < 0.05)	Small sample size No clinical outcome 10/18 0.9% saline trials were stopped early due to increases in ICP during infusion 6/9 patients in the normal saline arm crossed over
Simma et al. 1998 [5]	A prospective, randomized, and controlled study of fluid management in children with severe head injury: Lactated Ringer’s solution vs hypertonic saline	n = 32	Mean ICP was not different between patients who received 1.5% HTS and lactated Ringer’s (p > 0.05)	Patients in the lactated Ringer’s arm required more interventions for control of ICP (1047 vs 662, p < 0.05) Mortality was similar between groups (11.7% in lactated Ringers, 0% in HTS group)	Small sample size 1.5% hypertonic saline is less commonly used today Non-clinical primary outcome
Vialet et al. 2003 [6]	Isovolume hypertonic solutes (sodium chloride or mannitol) in the treatment of refractory posttraumatic intracranial hypertension: 2 mL/kg 7.5% saline is more effective than 2 mL/kg 20% mannitol.	n = 20	Episodes of ICP > 25 mmHg were significantly greater in 20% HTS than 7.5% HTS (13.3 episodes/day vs 6.8 episodes/day respectively, p < 0.05)	Total duration of ICP > 25 mmHg was significantly longer in patients treated with mannitol (95 min/day vs 62 min/day respectively, p < 0.05) Treatment failure was significantly more likely in mannitol-treated patients (70% vs 10%, p < 0.05) 3-month GOS was not different between groups (p > 0.05)	Small sample size Non-clinical primary outcome
Francony et al. 2008 [7]	Equimolar doses of mannitol and hypertonic saline in the treatment of increased intracranial pressure.	n = 20	Changes in ICP over the first 2 h were not different between 20% mannitol or 7.45% HTS (45% vs 32% respectively, p > 0.05)	No other ICP metrics were assessed	Small sample size Non-clinical primary outcome Assessed efficacy of single dose of hyperosmotic agent 3/20 patients were not TBI patients
Sakellaridis et al. 2011 [8]	Comparison of mannitol and hypertonic saline in the treatment of severe brain injuries	n = 29	Mean ↓ in ICP between 20% mannitol and 15% HTS was not different (7.96 vs 8.43 mmHg respectively, p > 0.05)	Mean duration of effect was similar between mannitol and HTS (p > 0.05) 3-month GOS was not different between groups (p > 0.05)	Small sample size Non-clinical primary outcome Initial treatment was randomized, but patients were subsequently alternated between therapies
Cottenceau et al. 2011 [9]	Comparison of effects of equiosmolar doses of mannitol and hypertonic saline on cerebral blood flow and metabolism in traumatic brain injury	n = 47	There was no difference between 20% mannitol and 7.5% HTS with respect to 6-month GOS scores (p > 0.05)	Proportional ↓ in ICP was greater with HTS than mannitol (p value not reported) The difference in the amount of time with ICP > 20 mmHg was not different between groups (11.1 h for HTS, 8.4 h for mannitol, p > 0.05)	Small sample size
Jagannatha et al. 2016 [10]	An equiosmolar study on early intracranial physiology and long term outcome in severe traumatic brain injury comparing mannitol and hypertonic saline	n = 38	There was no difference between changes in ICP over the first 6 days post injury between 20% mannitol or 3% HTS (p > 0.05)	There was no difference between groups with regard to proportion of time with ICP values < 20 mmHg (p > 0.05) ↓ in ICP following infusion was not different between groups (− 8.9 mmHg for mannitol and − 10.1 mmHg for HTS, p > 0.05) 6-month GOS was not different between groups (0% mannitol, 14% HTS, p > 0.05)	Small sample size Non-clinical primary outcome
Kumar et al. 2019 [11]	Comparison of equiosmolar dose of hyperosmolar agents in reducing intracranial pressure—a randomized control study in pediatric traumatic brain injury	n = 30	Mean ICP ↓ was not different between mannitol and HTS (− 7.13 mmHg vs − 5.67 mmHg, respectively)	No difference in mean ICP (p > 0.05) No difference in number of raised ICP episodes (p > 0.05)	Small sample size Non-clinical primary outcome
Patil and Gupta 2019 [12]	A Comparative Study of Bolus Dose of Hypertonic Saline, Mannitol, and Mannitol Plus Glycerol Combination in Patients with Severe Traumatic Brain Injury	n = 120	HTS, 20% mannitol, and 10% mannitol plus 10% glycerol were all able to reduce ICP to below 15 mmHg within an hour of infusion (p < 0.05)	HTS resulted in the largest percent change in ICP (60%, 55%, 57%, respectively, no statistics provided) HTS decreased ICP the fastest (16 min vs 23 min vs 19 min respectively, statistics not provided)	Statistical tests not reported for between-group comparisons Non-clinical primary outcome
Roquilly et al. ongoing [13] NCT03143751	Continuous hyperosmolar therapy for traumatic Brain-Injured patients (COBI)	n ~ 370	Effect of continuous infusion of 20% HTS on 6-month GOS-E	ICP control	Study currently recruiting

Favorable GOS: GOS > 3; Favorable GOS-E: GOS-E> 4; landmark studies in italics

GOS Glasgow outcome scale, GOS-E Glasgow outcome scale-extended, ICP intracranial pressure, PbtO₂ brain partial pressure of oxygen, HTS hypertonic saline

Sedation

Effects of sedatives on ICP were first noted in 1936 with hexobarbital [135]. Sedation use expanded with the realization that barbiturates could reduce the need for hyperosmotic agents and potentially improve outcomes [135–137]. In the 1980s, three RCTs (Table 3) evaluated barbiturates for ICP control without evidence of benefit. Current Brain Trauma Foundation guidelines provide a class IIb recommendation that barbiturates only be used for ICP control in patients who have failed both maximal medical management and surgery [64].

Table 3.

Randomized controlled trials of barbiturate coma after TBI

Authors (year)	Study title	Study size	Primary outcome result	ICP result	Study limitations
Schwartz et al. 1984 [14]	The University of Toronto head injury treatment study: a prospective, randomized comparison of pentobarbital and mannitol	n = 59	Mortality in patients randomized to pentobarbital was not different mannitol in patients with evacuated hematomas (p > 0.05) Mortality in patients randomized to pentobarbital was higher than mannitol in patients without an evacuated hematoma (70% vs 41% respectively, p < 0.05)	Pentobarbital was associated with higher rate of treatment failure than mannitol (rate not reported, p < 0.05)	Cross-over was differential, with more patients crossing from pentobarbital to mannitol Analysis of patients with hematoma is presented separately from those without hematoma, with no global analysis
Ward et al. 1985 [15]	Failure of prophylactic barbiturate coma in the treatment of severe head injury	n = 53	1-year GOS scores were not y different between patients treated with prophylactic pentobarbital vs standard of care (p > .05)	Mean ICP was not lower in patients treated with prophylactic pentobarbital when compared to standard of care (18.5 vs 19.5 mmHg respectively, p > 0.05)	Small sample size Hypotension occurred significantly more frequently in the pentobarbital group (p < 0.05) Barbiturate coma induction was not dependent on development of intracranial hypertension 5/27 patients in the pentobarbital arm were kept at subtherapeutic drug levels due to hypotension
Eisenberg et al. 1988 [16]	High-dose barbiturate control of elevated intracranial pressure in patients with severe head injury	n = 73	Control of intracranial hypertension with barbiturate therapy was not improved compared to standard care (32.4% vs 16.7% respectively, p > 0.05)	Treatment with barbiturates was associated with improved ICP control in a multivariate logistic regression model (ORs not provided, p < 0.05)	Significant cross-over between treatment arms prevented assessment of mortality A higher proportion of patients in the standard care arm had > 5 mm of midline shift on admission head CT (41.7% vs 18.9%, p not provided)

Favorable GOS: GOS > 3; Favorable GOS-E: GOS-E > 4; landmark studies in italics

GOS Glasgow outcome scale, ICP intracranial pressure

Non-barbiturate sedation following TBI is common, but not extensively studied. Most studies have compared propofol to midazolam, or the effect of various opioid agents on ICP (Table 4). Current guidelines recommend propofol sedation for ICP control [64]. While most studied in fentanyl, genetic factors may play a role in determining patient response to sedation dose and class [138–143]. Shein et al. recently compared bolus fentanyl to hypertonic saline for treatment of acute ICP spikes in pediatric severe TBI and demonstrated superiority of osmolar therapy [133].

Table 4.

Randomized controlled trials of non-barbiturate sedatives after TBI

Authors (year)	Study title	Sedatives compared	Study size	Primary outcome result	ICP result	Study limitations
Dearden and McDowall 1985 [17]	Comparison of etomidate and althesin in the reduction of increased ICP after head injury.	Etomidate vs althesin—all patients were receiving neuromuscular blockade	n = 10	1-year GOS was not different (p > 0.05)	Etomidate ↓, ICP between 10 and 30 min (p < 0.05), althesin ↓, ICP between 30 and 70 min after infusion (p < 0.05) Changes in ICP were not different by arm (p > 0.05)	Small sample size Adverse events not reported
Sperry et al. 1992 [18]	Fentanyl and sufentanil increase intracranial pressure in head trauma patients.	Fentanyl vs sufentanil	n = 9	Mean ↑ in ICP 7.7 (fentanyl) vs 6.1 mmHg (sufentanil) (p < 0.05)	Differences between drug with regard to ICP were not statistically significant 3/9 patients required hyperventilation (group unknown)	Small sample size No clinical outcome 7/9 patients received midazolam One patient required osmotherapy (intervention arm not described)
Levy et al. 1995 [19]	Propylene glycol toxicity following continuous etomidate infusion for the control of refractory cerebral edema.	Etomidate vs pentobarbital	n = 7	The study was stopped after enrollment of 7 patients due to AKI from propylene glycol toxicity	Post-treatment change in ICP was not different between groups (12 mmHg in etomidate vs 9 mmHg in pentobarbital) No patient required decompressive surgery	Small sample size AKI was likely due to the medication carrier. All patients received pancuronium Patients were sedated if increased ICP was refractory to mannitol, CSF drainage, and hyperventilation
Lauer et al. 1997 [20]	Opioid sedation does not alter intracranial pressure in head injured patients	Fentanyl vs sufentanil vs morphine; all had midazolam cointervention	n = 15	ICP did not change in 4 h in any of the groups (all p > 0.05)	No other markers of ICP or cerebral edema were addressed	Small sample size All patients were receiving mannitol for serum osmolality maintenance No clinical outcomes reported
Sanchez-Izquierdo-Riera et al. 1998 [21••]	Propofol vs midazolam: safety and efficacy for sedating the severe trauma patient.	Propofol and morphine vs midazolam and morphine vs propofol, midazolam, and morphine	n = 100	All treatment arms were equally effective for inducing and maintaining sedation (p > 0.05)	Mean ICP for propofol, midazolam, or both was not different between groups (16 mmHg, 16 mmHg, 14 mmHg, respectively, p > 0.05)	Only 59/100 patients were TBI patients, and only 40/100 had ICP monitoring The need for additional ICP controlling therapies between groups was not addressed
Kelly et al. 1999 [22]	Propofol in the treatment of moderate and severe head injury: a randomized, prospective double-blinded pilot trial.	Propofol plus morphine vs morphine	n = 42	No difference in 6-month favorable GOS score in propofolvs morphine (52.1% vs 47.4%, p > 0.05)	Mean ICP was lower in propofol-treated patients only on day 3 Benzodiazepines, neuromuscular blockade, and barbiturate requirements were higher in the morphine group (p < 0.05) CSF drainage was higher in the morphine group on days 2 and 3 (p < 0.05)	Small sample size Patients treated with propofol were more likely to be over 55, and more likely to have an enrollment GCS of 3–5 Patients treated with propofol were more likely to have effaced cisterns on admission
de Nadal et al. 2000 [24]	Cerebral hemodynamic effects of morphine and fentanyl in patients with severe head injury: absence of correlation to cerebral autoregulation.	1-mininfusion of morphine vs 1 fentanyl—cointerventions included vecuronium and midazolam	n = 30	Neither morphine nor fentanyl was associated with changes in cerebral blood flow	ICP was ↑ at both 5 min and 60 min following morphine and fentanyl boluses (p < 0.05) Change in ICP was not different between medications	Small sample size Planned cross-over prevents any outcome assessment between sedatives Cointervention confounds specificity of action
Sandiumenge Camps et al. 2000 [25]	Midazolam and 2% propofol in long-term sedation of traumatized, critically ill patients: Efficacy and safety comparison	Propofol plus morphine vs midazolam plus morphine	n = 63	ICU length of stay and ICU mortality were not different between groups (p > 0.05)	Mean ICP did not differ between groups (15 mmHg for midazolam and 14 mmHg for propofol, p > 0.05) ICP treatment intensity did not differ between groups (p > 0.05)	Limited sample size 27% of patients were not TBI Head injury characteristics not reported Long-term outcomes not reported
Bourgoin et al. 2003 [26]	Safety of sedation with ketamine in severe head injury patients: comparison with sufentanil.	Sufentanil plus midazolam vs ketamine plus midazolam	n = 12	Daily mean ICP and CPP were not different over the first 4 days (p > 0.05)	Daily ICP > 25 were not different over the first 4 days (p > 0.05) Need for other interventions for refractory ICP was not different (p > 0.05)	Small sample size No clinical outcome Cointervention confounds specificity of treatment
Bourgoin et al. 2005 [27]	Effects of sufentanil or ketamine administered in target-controlled infusion on the cerebral hemo -dynamics of severely brain-injured patients.	Sufentanil plus midazolam vs ketamine plus midazolam	n = 30	Sufentanil or ketamine dose was not associated with changes in ICP, CPP, or MAP (p > 0.05)	Other measurements of ICP control or edema were not reported	Small sample size No clinical outcome reported Cointervention confounds specific of intervention
Ghori et al. 2007 [28]	Effect of midazolam vs propofol sedation on markers of neurological injury and outcome after isolated severe head injury: a pilot study	Midazolam vs propofol	n = 28	No difference in good outcome (3-month GOS) between midazolam and propofol (54% vs 53%)	Daily mean ICP was not different between groups on days 0–3 (values not provided, p > 0.05) Propofol group was more likely to require decompressive surgery (p < 0.05), no difference in mannitol requirement (p > 0.05)	Small sample size Patients treated with midazolam were significantly more likely to have multiple contusions on head CT
Tanguy et al. 2012 [29]	Cerebral Microdialysis Effects of Propofol vs Midazolam in Severe Traumatic Brain Injury	Midazolam vs propofol	n = 25	No differences in microdialysis parameters between sedatives	ICP was not different (p > 0.05) No difference in the need for mannitol or decompressive surgery (p > 0.05)	Small sample size No clinical outcome reported

Favorable GOS: GOS > 3; landmark studies in italics

GOS Glasgow outcome scale, ICP intracranial pressure, CPP cerebral perfusion pressure, MAP mean arterial pressure, ICU intensive care unit

Hypothermia

Therapeutic hypothermia for neuroprotection was first documented in 1897 by Charles Phelps, neurosurgeon for the New York City police department. He described using the “ice cap,” indicating that it was second only to trephination in effectiveness when managing acute brain injury [144]. Focal hypothermia was used by Temple Fay in 1938 [145, 146] with implantation of an “ice bomb” into the residual cavity after intracranial tumor resection [146], and subsequently in TBI, with clinical improvement [146]. Soon thereafter, Nils Lundberg demonstrated that hypothermia was associated with a reduction in ICP in his seminal 1960 paper on monitoring ICP [94].

Interest in systemic hypothermia after TBI has cooled after unexpectedly disappointing results from modern RCTs including Eurotherm3235 and POLAR with regard to ICP control and functional outcome (Table 5) [147••, 148]. This has been redirected towards selective cooling of the head [149, 150]—hearkening back to the work of Drs. Phelps and Fay. Nonetheless, exciting results from research into cold-shock proteins, cold-stress hormones, and ultra-mild hypothermia may ultimately “decode” molecularly targeted benefits of this process and re-invigorate clinical investigations [151]. Current guidelines do not recommend the use of prophylactic hypothermia for ICP control after TBI, although it does represent a second-tier therapy for refractory intracranial hypertension.

Table 5.

Randomized controlled trials of therapeutic hypothermia after TBI

Authors (year)	Study title	Duration of hypothermia and target temperature	Study size	Primary outcome result	ICP result	Study limitations
Marion et al. 1993 [30]	The use of moderate therapeutic hypothermia for patients with severe head injuries: a preliminary report.	24 h, 32–33 °C	n = 40	No difference in favorabLe 3-month GOS score between hypothermia and normothermia (60% vs 40%, p > 0.05)	ICP was lower with hypothermia during the first 24 h (p < 0.05) Proportion of ICP >20 mmHg was lower with hypothermia over 36 h (13% vs 25%, p < 0.05) Therapeutic intensity was not different with hypothermia during cooling, but was higher after rewarming (p < 0.05)	Small sample size Statistical tests did not properly account for time-variant data Limited data on enrollment and loss to follow-up
Clifton et al. 1993 [31]	A phase II study of moderate hypothermia in severe brain injury.	48 h, 32–33 °C	n = 46	No difference in favorable 3-month GOS score with hypothermia (52.2%) vs normothermia (36.4%, p > 0.05)	There were no differences in ICP during the first 72 h of admission (p > 0.05 at all points) Therapy intensity level was higher with hypothermia (p < 0.05)	Small sample size Limited information on enrollment and loss to follow-up
Shiozaki et al. 1993 [32]	Effect of mild hypothermia on uncontrollable intracraniaL hypertension after severe head injury.	48 h, 33.5–34.5 °C	n = 33	Favorable 6-month GOS was not different between hypothermia and normothermia	Mean ICP ↓ by 10.4 mmHg after hypothermia (p < 0.01). Not compared to normothermia 6-month mortality less likely with hypothermia (31% vs 71%, p < 0.05)	Small sample size Study limited to patients who failed all other methods of ICP control Mean ICP prior to randomization was 36.9 mmHg in control and 35.4 mmHg in hypothermia
Marion et al. 1997 [33]	Treatment of Traumatic Brain Injury with Moderate Hypothermia	24 h, 32–33 °C	n = 82	Favorable 3-month GOS score was more likely with hypothermia vs normothermia (38% vs 17%, p < 0.05)	Mean ICP was lower with hypothermia from 0 to 36 h (15.4 vs 19.7 mmHg, p < 0.05) Mean ICP was not different by 37–60 h (17.4 vs 19.2 mmHg, p > 0.05)	Small sample size Limited information on why only 155/82 eligible patients were randomized Limited information on loss to follow-up
Shiozaki et al. 1999 [34]	Little benefit from mild hypothermia therapy for severely head injured patients with low intracranial pressure	48 h, 33.5–34.5 °C	n = 16	6-month favorable GOS score equally Likely in hypothermia and normothermia (75% vs 87.5%, p > 0.05)	Mean ICP was not different between hypothermia and normothermia groups (10.9 vs 10.8 mmHg, p > 0.05)	Small sample size Patients whose ICP could not be maintained less than 20 mmHg were excluded from the study
Jiang et al. 2000 [35]	Effect of long-term mild hypothermia therapy in patients with severe TBI: 1-year follow-up review of 87 cases.	72–336 h, 33–35 °C	n = 87	1-year favorable GOS more likely with hypothermia vs normothermia (46.5% vs 27.3%, p < 0.05)	ICP was not lower with hypothermia at 24 h (29.6 vs 30.3 mmHg, p > 0.05) ICP was lower in the hypothermia arm at 7 days (18.9 vs 28.1 mmHg, p < 0.05)	Small sample size Limited information on patient enrollment and loss to follow-up Complications only reported at 1 year All patients were receiving mannitol Q6-8 h
Clifton et al. 2001 [36]	Lack of Effect of Induction of Hypothermia after Acute Brain Injury (NABIS:H)	48 h, 32.5–34 °C	n = 385	6-month favorabLe GOS was 43% in both groups (p > 0.05)	Mean ICP did not differ on days 1–4 between groups (p > 0.05) Proportion of patients with ICP vaLue over 30 was lower in hypothermia than normothermia (41% vs 59%, p < 0.05) Therapy intensity was similar between groups (p > 0.05)	Hypotension-associated organ dysfunction is was more common in the hypothermia arm (10% vs 3%, p < 0.05) Limited information on patient enrollment and loss to follow-up 35% of patients in normothermia arm were hypothermic during first 16 h
Gal et al. 2002 [37]	Mild hypothermia therapy for patients with severe brain injury	72 h, 34 °C	n = 30	6-month favorable GOS was not different (80% hypothermia, 47% normothermia, p > 0.05)	Mean ICP was lower in hypothermia than normothermia (12 vs 18 mmHg, p < 0.05)	Small sample size Insufficient detail on other treatments for ICP
Zhi et al. 2003 [38]	Study on therapeutic mechanism and clinical effect of mild hypothermia in patients with severe head injury	24–168 h, 32–35 °C	n = 396	6-month GOS = 5 was higher in hypothermia than normothermia (38.8% vs 19%, p < 0.05)	Mean ICP was lower at 24, 72, and 196 h in hypothermia than normothermia (21.4, 14.8, 14.0 mmHg vs 25.5, 22.0, 20.6 mmHg, p < 0.05 for all comparisons)	Limited information on patient enrollment and follow-up No information on blinding Limited information of treatment-related complications
Hashiguchi et al. 2003 [39]	Mild hypothermia reduces expression of heat shock protein 60 in leukocytes from severely head-injured patients.	48 h, 34 °C	n = 17	Heat-shock protein expression was lower in hypothermia throughout the first 14 days post-injury (all p < 0.05)	Favorable 6-month GOS was not significantly difference between hypothermia and normothermia (67% vs 87.5%, p > 0.05) No ICP data was reported	Small sample size Non-clinical primary outcome ICP results not reported
Smrcka et al. 2005 [40]	The influence of mild hypothermia on ICP, CPP, and outcome in patients with primary and secondary brain injury	72 h, 32–35 °C	n = 72	Mean 6-month GOS was higher in hypothermia than normothermia arm (4.1 vs 3.2, p < 0.05)	Mean ICP was lower in the hypothermia arm than the hypothermia arm (11.8 vs 17.7 mmHg, p < 0.05)	Limited information on enrollment and loss to follow-up Small sample size Inappropriate statistical test of GOS No presentation of complications of hypothermia
Liu et al. 2006 [41]	Effects of selective brain cooling in patients with severe traumatic brain injury: a preliminary study.	72 h, 33–35 °C, A: selective head and neck cooling, B: systemic cooling, C: normothermia	n = 66	2-year favorabLe GOS was more common in hypothermia than normothermia arm (A: 72.7%, B: 57.1%, C: 34.8%)	ICP was lower in hypothermia groups than normothermia at 24, 48, and 72 h (A: 27.4, 29.2, 26.7 mmHg, B: 27.1, 29.7, 26.5 mmHg, C: 32.6, 34.8, 31.8 mmHg, p < 0.05 between hypothermia arms and normothermia	Limited information on enrollment and loss to follow-up Small sample size per arm Blinding not described
Qui et al. 2007 [42]	Effects of therapeutic mild hypothermia on patients with severe TBI after craniotomy	96 h, 33–35 °C	n = 80	ICP at 24, 48, and 72 h were lower with hypothermia than normothermia (23.5, 24.7,22.5 mmHg vs 25.9, 25.9, 24.6 mmHg, all p < 0.05)	Favorable 1-year GOS score occurred more commonly in hypothermia arm (70% vs 47.5%, p < 0.05)	Study limited to post-operative patients Small sample size Limited information on patient enrollment and loss to follow-up Pneumonia more common in hypothermia arm
Lee et al. 2010 [43]	Applying cerebral hypothermia and brain oxygen monitoring in treating severe TBI.	Unclear duration, 33–35 °C—3 groups: ICP monitoring (A), hypothermia + ICP monitoring (B), hypothermia + PbtO2 monitoring (C)	n = 45	ICP on days 3–5 was lower in hypothermia than normothermia (p < 0.05 for all days)	Mean ICP was lower in the hypothermia arms than the normothermia arm (C: 16.0, B: 17.7, A: 20.4, global F test < 0.05, post hoc not reported) 6-month favorable GOS score was more likely in hypothermia than normothermia arm (C: 71.4%, B: 60%, A: 50%, global chi-squared test < 0.05)	Small numbers of patients in each group Limited information on patient enrollment and loss to follow-up Insufficient post hoc testing Statistical tests of complications not reported
Zhao et al. 2011 [44]	Mild hypothermia therapy reduces blood glucose and lactate and improves neurologic outcomes in patients with severe TBI.	72 h, 33 °C	n = 81	Blood glucose and lactate were lower 24 and 72 h in hypothermia vs normothermia (p < 0.05)	↓ mean ICP with hypothermia at 24 and 72 h (19.8, 18.7 vs 25.8, 27.4, all p < 0.05) Favorable 6-month GOS occurred in 75% of hypothermia patients and 51.2% of normothermia patients (p < 0.05)	Limited data on enrollment and loss to follow-up Need for other ICP treatments unclear Non-clinical primary outcome Complications of hypothermia not described
Clifton et al. 2011 [45]	Very early hypothermia induction in patients with severe brain injury (the National Acute Brain Injury Study: Hypothermia II): a randomised trial. (NABIS: H II)	48 h, 33–35 °C	n = 97	Favorable 6-month outcome was not different between hypothermia and normothermia (40% vs 44%, p < 0.05)	Number of episodes of increased ICP were higher in hypothermia group (335 vs 148, p < 0.05) Therapeutic intensity was higher the hypothermia arm (p < 0.05)	Limited data on why approximately 90% of patients were not enrolled Small sample size > 50% of patients randomized were excluded from study Small sample size
Andrews et al. 2015 [46]	Hypothermia for Intracranial Hypertension after TBI (Eurotherm3235)	>48 h, 32–35 °C	n = 387	Favorable 6-month GOS-E score lower with hypothermia vs normothermia (25.7% vs 36.5%, p < 0.05)	Mean daily ICP was not different at 7 days (p > 0.05) Normothermia arm more likely to require barbiturates for ICP control (21.7% vs 10.4%) No difference in surgery	Patients were enrolled prior to onset of intracranial hypertension Study terminated early due to worse outcomes in the interventional arm Concerns that differences in adverse events were due to lack of blinding
Maekawa et al. 2015 [47]	Prolonged mild therapeutic hypothermia vs fever control with tight hemodynamic monitoring and slow rewarming in patients with severe TBI: a randomized controlled trial.	> 72 h, 32–34 °C	n = 148	Favorable 6-month GOS score was not different between hypothermia and normothermia groups (47% vs 52%, p > 0.05)	Mean ICP on days 0,1, and 3 were not different between groups (p > 0.05 for all comparisons)	Study stopped early due to low enrollment Sedation protocol changed mid-study Injury severity score was lower in the normothermia group The need for additional therapies for ICP control was not reported 62% of patients in the hypothermia arm did not reach therapeutic temperature within 6 h per protocol
Tang et al. 2017 [48]	Mild induced hypothermia for patients with severe TBI after decompressive craniectomy	48 h, 33–35 °C	n = 60	Favorable 6-month GOS was not different between hypothermia and normothermia (40% and 36.5%, p > 0.05). 1 mortality with hypothermia (p < 0.05)	ICP was lower i7n the hypothermia group over the first 5 days of the study (p < 0.05)	Study limited to patients who had intracranial hypertension refractory to decompressive surgery Small sample size Hypothermia was associated with increased risk of pneumonia and electrolyte abnormalities
Cooper et al. 2018 [49]	Effect of Early Sustained Prophylactic Hypothermia on Neurologic Outcomes Among Patients With Severe TBI (POLAR)	72–168 h, 33–35 °C	n = 466	Favorable 6-month GOS-E score was not different in hypothermia vs normothermia (48.8% vs 49.1%, p > 0.05)	End-hourly ICP > 20 mmHg were not different in hypothermia vs normothermia (6 and 6, p > 0.05) Need for hyperosmotic or sedative agents for ICP control was not different (all p > 0.05) Need for decompressive surgery was not different (p > 0.05)	10.4% of patients randomized to the hypothermia did not reach 35 °C, and 28.5% never reached 33 °C 38.1% of patients in the hypothermia arm were withdrawn early from hypothermia, largely due to alcohol confounding the determination of severe TBI

Favorable GOS: GOS > 3; Favorable GOS-E: GOS-E > 4; landmark studies in italics

GOS Glasgow outcome scale, GOS-E Glasgow outcome scale-extended, ICP intracranial pressure, PbtO₂ brain partial pressure of oxygen

Treatments under investigation

Ongoing research efforts include focused investigations targeting molecular pathways underlying TBI edema. There is also a formidable body of research evaluating the related process of neuroinflammation (and systemic inflammation) reviewed elsewhere [15]. Targeting individual pathways is unlikely to be sufficient; multi-pronged approaches similar to chemotherapeutic regimens, either with drugs or novel molecular or cellular therapies, may be warranted, particularly given the overlap and cross-talk between mechanisms. The therapies outlined below are not exhaustive, but represent promising targets supported by preclinical research (Fig. 2), some of which are now being evaluated in human TBI.

Fig. 2.

This figure demonstrates physiologic contributors to cerebral edema and possible molecular targets for future therapies. It includes the intravascular space with red blood cells (RBCs), the vascular endothelium with an associated subglycocalyx space, and the interstitial space containing neuronal bodies in addition to other cells. Starling forces controlling edema development depend on hydrostatic pressures within the capillary (P_c) and interstitial space (P_i), and on oncotic pressures in the capillary (π_c), subglycocalyx space (π_sg), and interstitial space (π_i). Key molecular contributors include the sulfonylurea-1/transient receptor potential cation channel subfamily member-4 (Sur-1/Trpm-4), Na-K-Cl-cotransporter 1 (NKCC1), vasopressin receptor 1 (V1), and aquaporin-4 (AQP4).

Glibenclamide (glyburide)

Glibenclamide, a first-generation sulfonylurea, has been used for decades in type 2 diabetes mellitus [152, 153]. Glibenclamide binds to sulfonylurea receptor 1 (SUR1), a protein that regulates ion channels [38]. The SUR1-TRPM4 cation channel is not normally expressed in the CNS, but is upregulated in the neurovascular unit in a number of CNS pathologies [154••, 155–160]. In periods of relative or absolute cerebral ischemia like stroke or TBI, channel opening results in sodium influx, causing membrane depolarization, water influx (via associated AQP-4 channels), and further oncotic swelling [161]. This contributes to cytotoxic edema. In endothelial and astrocytic cells, this may contribute to BBB disruption and vasogenic edema [62, 160, 162]. Unchecked, this process can cause cell death [4, 154••, 160, 163]. Increased CSF SUR1 in severe TBI patients is associated with increased odds of cerebral edema [156], and SUR1 is overexpressed in pathologic tissue obtained from surgical evacuation of traumatic contusions [164]. Genetic variation in both ABCC8 (Sur-1 gene) and TRPM4 is associated with cerebral edema and intracranial hypertension after TBI, suggesting individual differences in this pathway [37–39, 67].

Glibenclamide binding to SUR1-TRPM4 increases the probability of the channel remaining in the closed configuration [4, 153]. In the CNS, this leads to decreased ion flux across cell membranes, disrupting edema formation and oncotic cell death [165–167]. Preclinical TBI models have shown that glibenclamide decreases cerebral edema, ICP, contusion expansion, and cell loss [155, 162, 163, 165, 168–170]. One RCT of 40 patients with diffuse axonal injuries reported improved outcomes with oral glibenclamide [171]. In another RCT of 66 moderate-severe TBI patients, oral glibenclamide was associated with decreased contusion expansion [172]. A phase II trial of intravenous glibenclamide reported no difference in lesion expansion and MRI measures of edema vs placebo, though the study was limited by small sample size (n = 28)[173••]. A larger phase II trial is ongoing (NCT03954041).

AER-271

AER-271 is a small molecule AQP4 inhibitor. AQP family proteins allow water to traverse bipolar plasma membranes, with AQP4 being the most highly expressed AQP in the brain, spinal cord, and optic nerve [32]. AQP4 appears to have an important, complex, role in the pathophysiology of cerebral edema following TBI [25, 174–177]. Changes in cellular tonicity, such as those that occur with ionic shifts preceding cytotoxic edema, may lead to rapid localization of AQP4 to the plasma membrane and facilitate cellular swelling [178–180]. However, depending on the time-point or model, decreased AQP4 levels are also associated with increased vasogenic edema [25, 177, 181]. The effects of AQP4 knockout or inhibition after preclinical TBI are conflicting and may depend on injury model, spatial, and temporal expression patterns [31, 174, 176, 181–191]. To our knowledge, there are no clinical studies of AER-271, or other AQP4 inhibitors, in TBI.

Bumetanide

NKCC1 is constitutively expressed on plasma membranes of multiple CNS cell types and is further upregulated within an hour of injury in preclinical TBI models [192, 193]. NKCC1 has been associated with cerebral edema development, and transporter activity is associated with neurotoxicity [193–195]. As with SUR1, ATP depletion leads to increased ion flux through the NKCC1 transporter: failure of the Na⁺/K⁺ pump results in increased extracellular potassium, driving uptake of chloride and sodium into the cell [62, 193]. The resultant transmembrane flux results in cellular swelling, contributing to cytotoxic edema [62, 196]. Increased NKCC1 activity following brain injury may also contribute to the upregulation of aquaporin proteins, further compounding cerebral edema generation [193].

Bumetanide is a potent loop diuretic that inhibits NKCC1 and is used primarily for treatment of systemic volume overload [197, 198]. In an in vitro TBI model, bumetanide reduced astrocyte swelling [196]. In vivo, bumetanide increases BBB integrity and decreases AQP4 expression, cellular swelling, and neuronal apoptosis in TBI and spinal cord injury models [195, 199, 200]. CNS bumetanide concentration may be limited by active drug efflux transporters, reducing its efficacy in pre-venting cerebral edema [201, 202]. There have been no human trials of bumetanide in TBI.

Vasopressin antagonists

Arginine vasopressin (AVP), or antidiuretic hormone, is a peptide synthesized by the hypothalamus [4]. The V1a receptor, constitutively expressed in the CNS and upregulated after TBI, may play a role in cerebral edema [52, 203, 204]. Both pharmacologic inhibition and genetic knockout of the V1a receptor are neuroprotective and associated with decreased cerebral edema [204–207]. The mechanism, though not fully elucidated, likely includes maintenance of BBB integrity and decreased AQP4 expression [203, 205]. Unexpectedly, studies of AVP as a vasopressor vs conventional use of catecholamines in human TBI found no evidence of worsening cerebral edema between groups [208, 209].

Nonetheless, in an RCT of ten severe TBI patients, a single dose of conivaptan (V2 and V1a receptor antagonist) lowered ICP and increased serum sodium at 4 h post-administration [210]. However, this also produced a lasting diuretic effect, indicating a need for close fluid status assessment [210]. In preclinical TBI, SR 49059 (small molecule inhibitor of V1a), and V1880 (peptide selectively inhibiting V1 receptors), reduced cytotoxic edema and ICP [205, 206]. There are no clinical studies of V1-selective antagonists, but this could help mitigate diuretic effects seen with conivaptan.

Conclusions

Although cerebral edema has long been recognized to contribute to unfavorable outcome after TBI, management strategies remain suboptimal and limited by side effect profiles, lack of treatment specificity, and unclear benefit on functional outcomes. This underscores the need for further investigation into key underlying pathophysiological processes which will impact both trial design and clinical care. Clinical practice of precision medicine in other fields is exponentially growing to include endophenotyping coupled to molecular targeting and point-of-care genetic testing to guide treatment [211–216]. Complementing molecular, neuromonitoring, and imaging biomarkers with genetic profiles to stratify TBI edema may result in clinical trials that identify patients most likely to benefit from specific interventions—thereby facilitating the transition of precision medicine-guided therapy in TBI from an elusive ideal to a clinical reality.