Spontaneous intracerebral hemorrhage (ICH) accounts for approximately 20% of all strokes and affects about 3 million patients worldwide each year (GBD 2019 Stroke Collaborators, 2019). ICH is devastating, with a mortality of ~40% at 1 month, and more than two-thirds of survivors remain dependent, requiring long-term care (van Asch et al., 2010; Krishnamurthy et al., 2020). Compared to advances in the management of acute ischemic stroke, the treatment for ICH lags significantly. Poor prognosis is associated with the hematoma size, location, and intraventricular hemorrhage. Hence, randomized clinical trials have assessed limiting the hemorrhage size and expansion through hemostasis, lowering blood pressure (BP), and surgery (Broderick et al., 1993; Mendelow et al., 2005, 2013; Anderson et al., 2008; Investigators, 2015; NCT02880878, 2023). Except for surgical evacuation of a hematoma in a highly selected group of patients (Ratcliff et al., 2023), no other intervention has been shown to improve functional outcomes. Therefore, the management of ICH remains largely supportive. Survivors of ICH are at risk of death or severe disability due to the cerebral edema that surrounds the hematoma (perihematomal edema [PHE]) (Venkatasubramanian et al., 2011; Balami and Buchan, 2012; Yang J. et al., 2015; Hurford et al., 2019). In a large-volume ICH, the accompanying swelling can increase the mass effect and lead to brain herniation (Zazulia et al., 1999). Hence, PHE is emerging as a key component of secondary brain injury and a potential surrogate outcome measure for preclinical and clinical trials (Venkatasubramanian et al., 2011; Yang Z. et al., 2015). Understanding the mechanisms and natural history of PHE is therefore important, and the development of potential therapeutic agents is of interest to clinical trialists and researchers.

The pathogenesis of PHE and accompanying brain injury are not yet fully understood but include vasogenic, cytotoxic, inflammatory, and oxidative mechanisms accompanied by disruption of the blood-brain barrier (BBB) (Figure 1). In this narrative review, we focus on these mechanisms, the natural history of PHE, neuroimaging measures, and laboratory parameters that may be associated with the pathogenesis, and summarize potential interventions, including data from translational and clinical studies. We also list the gaps in the current knowledge and suggest directions for future trials of cerebral edema that could potentially change clinical practice.

The precise natural history of PHE is unclear, but it is thought to evolve over three phases.

The immediate response to intracerebral hemorrhage is activation of the coagulation cascade to limit bleeding and here, thrombin has an important role. However, experimental studies have shown that in high concentrations as in acute stroke, thrombin induces edema independent of fibrinogen and causes the release of nitric oxide and cytokines, including tumor necrosis factor–alpha (TNF-α), interleukin (IL)-12, and IL-6 (Xi et al., 1998, 2003). In addition, thrombin promotes recruitment and infiltration of neutrophils, lymphocytes, and macrophages into the perihematomal region by inducing chemotactic substances and adhesion molecules (Xi et al., 2003). Thrombin also directly stimulates microglia through specific proteinase-activated receptors (PARs) (Chen et al., 2022).

The activation of thrombin can cause an opening of the BBB through PARs and damage the basement membrane by activating metalloproteinases (MMPs) (Chen et al., 2022). MMPs are proteolytic enzymes, which degrade components of the basement membrane, including collagen, gelatin, laminin, and fibronectin. Thrombin can also trigger activation of Src kinase (a proto-oncogene protein with kinase activity), which, in turn, potentiates damage to astrocytes and endothelial cells, the main cellular component of the BBB (Xi et al., 2003).

In response to thrombin activation and damage to the BBB, leukocytes enter the perihematomal region. Studies have shown that the activation of a specific receptor, toll-like receptor 4 (TLR-4), on the surface of neutrophils and macrophages is important for infiltration (Yang J. et al., 2015). The activation of TLR-4 begins within hours after a stroke and continues for about a week (Yang J. et al., 2015). By inducing the expression of a transcription factor, nuclear factor–kappa B (NF-κB), on microglia and macrophages, TLR-4 signaling also stimulates the release of pro-inflammatory cytokines (e.g., IL-1, IL-6, TNF-α, and IL-β), signaling pathways generating free radicals, and glutamate (Sansing et al., 2007; Yang J. et al., 2015).

Neutrophils are the earliest to reach the perihematomal region, and studies suggest that the peak activity is at ~72 h. Neutrophils seem to contribute to cerebral edema by generating reactive oxygen species (ROS) and pro-inflammatory protease enzymes that damage the BBB (Wang and Dore, 2007). The activation of microglia occurs later, and studies have shown this to continue for ~2 weeks (Zheng et al., 2020). The inflammation mediated by microglia is driven by its M1 phenotype through the release of cytokines, chemokines, ROS, and nitric oxide (Wang and Dore, 2007). The activity of the M1 phenotype is maintained by activated astrocytes and lymphocytes (T1 helper cells) (Tschoe et al., 2020). At ~1 week of ictus, the M1 phenotype converts to the M2 phenotype, and this begins the process of resolution of inflammation, scar formation, and brain repair (Wu J. et al., 2017). The conversion from the M1 to the M2 phenotype is potentiated by anti-inflammatory cytokines (IL-4, IL-10, IL-13), lymphocyte T2 helper cells and a number of transcription factors (Chhor et al., 2013). Hence, the effects of microglia after ICH are either pro-inflammatory or neuroprotective and time-dependent. The anti-inflammatory effect of the M2 phenotype has led to testing of potential treatments, which is discussed later.

The disruption to the BBB also leads to the activation of the complement system in the perihematomal region (Lee et al., 1995). Both complement C3a and C5a are anaphylatoxins and induce endothelial cells and infiltration of pro-inflammatory cells (Lee et al., 1995; Holste et al., 2021). Complement activation also leads to the formation of the membrane attack complex (MAC), which causes erythrolysis. MAC is known to be directly toxic to neurons, glial cells, astrocytes, and the BBB (Holste et al., 2021).

Although erythrolysis begins early after ICH, the late phase of edema is mainly medicated by the release of toxic products, including hemoglobin and iron, and phagocytosis (Ironside et al., 2019).

Hemoglobin has been shown to inhibit the enzyme Na+K+ATPase and induce the release of ROS and peroxidation of membrane lipids, leading to neuronal death (Bautista et al., 2021). Haem has been shown to activate microglia through TLR-4, and this, in turn, can induce NF-κB through a complex system of gene expression or signaling pathways (Tschoe et al., 2020). This process leads to more release of inflammatory cytokines and brain injury. (Tschoe et al., 2020).

Following the breakdown of the hematoma, iron is first observed in the perihematomal region at day 1, reaching peak level at approximately 7 days, and can continue to remain at that level for approximately 2 weeks (Qing et al., 2009). There is significant evidence that excess iron leads to brain damage (Xi et al., 2002). One mechanism is through the release of toxic free radicals (“ferroptosis”) (Bautista et al., 2021). Experimental studies have also shown that iron induces inflammation by generating ROS and MMP-9 and BBB dysfunction and increases cytotoxic swelling by activating AQP4 (Qing et al., 2009).

Phagocytosis leads to brain repair and is mediated by the infiltrated microglia and macrophages in the perihematomal region (Galloway et al., 2019). Phagocytosis includes the following steps: anchoring, internalization, and, finally, processing of the products of hematoma breakdown (Hochreiter-Hufford and Ravichandran, 2013). The process itself is complex and regulated by multiple receptors and enzymes that are present on the surface of the microglia and macrophage itself (Hochreiter-Hufford and Ravichandran, 2013; Galloway et al., 2019). As phagocytosis continues, ROS and toxic free radicals are generated, which can lead to more neurotoxicity (Hu et al., 2016). Preclinical studies have shown that accelerating phagocytosis could limit the period of exposure of the brain to toxic products and lead to recovery (Zhao et al., 2007a, 2015a).

It is suggested that, after ICH, excess levels of the antidiuretic hormone vasopressin induce edema and inflammation by activating astrocytes (Hertz et al., 2014). Evidence also indicates that vasopressin increases BBB permeability and water accumulation through the expression of AQP4 (Hertz et al., 2014; Zhao et al., 2015b).

Recent work has shown that endothelin-1, a hormone that is released by endothelial cells, increases within hours after stroke and correlates with BBB hyperpermeability (Li et al., 2022). In addition, endothelin-1 is linked to genes that regulate the secretion of cytokines, inflammation, the release of toxic free radicals, and iron metabolism (Wang et al., 2013).

An in vivo microdialysis study showed that apart from glutamate, there is an excessive extracellular accumulation of amino acids, including taurine, glycine, and asparagine in acute ICH (Shuaib et al., 2003). The role of these amino acids in the pathogenesis of inflammation and prognosis is unclear and needs further exploration (Kanthan et al., 1995; Shuaib et al., 2003).

Following the third phase of cerebral edema, some studies suggest a gradual decline in volume while others suggest otherwise. One study found that edema volume at 1 month was similar to that observed early after ictus (Fung et al., 2016). Peng et al. showed that edema volume at 2–3 weeks was much higher than baseline and was associated with poor outcomes (Peng et al., 2019). Another observational study reported that PHE can last up to 2 months (Chen et al., 2021).

BP is elevated in ~70% of patients with acute stroke and is associated with poor outcomes (Leonardi-Bee et al., 2002). Following the loss of cerebral autoregulation, high BP in ICH is associated with increased rebleeding, hematoma expansion, and PHE (Willmot et al., 2004). A history of high BP is also relevant as shown by analysis from the Intensive Blood Pressure Reduction in Acute Cerebral Hemorrhage (INTERACT) trial, which showed a significant association with an increase in edema at 72 h compared to baseline (Arima et al., 2009), In addition, BP variability is linked to the pathogenesis of edema in the hyperacute and subacute phases, but the mechanisms are not fully understood (Sykora et al., 2009; Havenon et al., 2018). One explanation is that the variation in BP directly increases the hydrostatic pressure and oncotic pressure in the PHE region (Sykora et al., 2009). Another explanation is that significant fluctuation in BP induces the release of pro-inflammatory cytokines, hyperglycemia, BBB dysfunction, and vasogenic edema (Sykora et al., 2009).

To date, no studies have prospectively tested the effects of BP lowering on PHE, but post-hoc analyses of two trials, INTERACT-2 and the Antihypertensive Treatment in Acute intracerebral hemorrhage (ATACH-2) trial, suggest that the rate of edema expansion was less in participants randomized to intensive treatment (Yang J. et al., 2015; Leasure et al., 2019). It is noteworthy that the majority of participants in INTERACT-2 and ATACH-2 had deep ICH, and one reason for the observed result could be that there is a difference in hematoma dynamics when compared to other brain locations (Seiffge et al., 2022). The trials excluded patients presenting in later-time windows, with large ICH volume and very high systolic BP (>200 mm Hg), and so the pathogenesis of PHE in such patients needs further research (Anderson et al., 2013; Bath et al., 2019).

Studies have used various measures of PHE to assess the effect on prognosis. Some studies have used absolute PHE volume in the first hours after stroke and others have suggested the change from baseline to be useful (Sansing et al., 2003; Appelboom et al., 2013; Yang J. et al., 2015). Some authors have postulated using the ratio of PHE per mL of ICH size/volume (relative edema), but it can be difficult to estimate the effects if the hematoma itself is large (Staykov et al., 2011; Appelboom et al., 2013).

Recent work has proposed that the absolute or relative edema change per hour could be more useful as it reflects the actual speed at which the pathogenesis occurs, but this needs to be tested in larger studies (Grunwald et al., 2016).

It may be that examining the peak/highest value of PHE is more appropriate, but performing sequential scans in unwell ICH patients may not be practical (Staykov et al., 2011; Venkatasubramanian et al., 2011). However, the change in PHE is dependent on the hematocrit, which is the proportion of blood volume occupying an erythrocyte, and is known to be higher in males (Venkatasubramanian et al., 2011). A higher hematocrit could therefore expose the brain to more products of erythrolysis over time, and this could delay the peaking of edema. By comparison, in intraventricular hemorrhage, the peak value of edema can be reached more quickly as the hematocrit is diluted by the volume of cerebrospinal fluid (Venkatasubramanian et al., 2011).

Because the development of PHE is linked to ICH volume, it is difficult to assess the independent effects in the first few hours of stroke. Parry-Jones et al. suggested that edema extension distance (EED) or measuring the change in OED over time might be able to overcome this issue (Parry-Jones et al., 2015). OED is the linear measure of the extension of swelling from the boundary of the hematoma, and so theoretically, it should be independent of the hematoma. However, OED is based on the assumption that a hematoma is elliptical in shape and so may not be applicable to irregular ICH (Parry-Jones et al., 2015).

Studies have shown that high serum glucose at baseline is linked with the pathogenesis of PHE (Esposito et al., 2002; Qureshi et al., 2011). Hyperglycemia has been shown to mediate this effect through the release of TNF-α and IL-1 (Esposito et al., 2002). Other laboratory measures, including a high red blood cell count, increased platelet count, and increased hematocrit, are associated with an increase in PHE in the first week after symptom onset (Sansing et al., 2003; Venkatasubramanian et al., 2011). It is suggested that an increase in thromboplastin time correlates with the expansion of PHE, which implies the activation of the coagulation system (Sansing et al., 2003). A meta-analysis including 2,176 patients analyzed the prognostic effect of neutrophils and lymphocytes on outcomes after ICH (Liu et al., 2019). The results showed that a high neutrophil-to-lymphocyte ratio predicted an increased risk of early death, death, or disability at day 90 (Liu et al., 2019). Pathophysiologically, this can be explained as ongoing inflammation in a patient with a weakened immune system who is likely to have a poor outcome.

With advances in understanding the natural history of PHE, there is interest in assessing the prognostic effects. The mechanisms are not yet understood, and one reason is because the pathogenesis of PHE is linked to the presenting hematoma, which could confound the clinical effects (Gebel et al., 2002; Arima et al., 2009; Wu J. et al., 2017). In patients with ICH volume <30 cm³, observational data have shown that the presence of PHE at baseline increases the risk of death or disability at day 90 (Appelboom et al., 2013; Murthy et al., 2015). One explanation is that in small hematomas, a modest increase in the mass effect from PHE could lead to clinical deterioration. However, such patients could be potentially eligible for testing interventions for edema, which, in turn, could improve outcomes.

After adjusting for ICH volume, studies have used other measures of PHE and assessed the effect on clinical outcomes. Gebel et al. reported that there was an inverse correlation between relative PHE and death or disability at day 90 but other authors found no association (Gebel et al., 2002; Arima et al., 2009; Staykov et al., 2011; Volbers et al., 2016). Another study examined EED and found that a large deviation induced by swelling increased the risk of midline shift and brain herniation (Wu J. et al., 2017). Recently, Urday et al. showed that the peak increase in PHE between baseline and 24 h was independently associated with death at day 90 (Urday et al., 2016). Studies have also attempted to assess the effects of PHE beyond the hyperacute phase and shown that an increase up to 72 h predicts poor functional outcomes (Sansing et al., 2011a; Li et al., 2013; Bakhshayesh et al., 2014; Yang J. et al., 2015; Urday et al., 2016).

Research suggests that the prognostic effects of PHE depend on the location of the stroke itself (Appelboom et al., 2013; Grunwald et al., 2016; Leasure et al., 2019). This could be highly relevant in deep hematomas in the thalamus or basal ganglia, as these structures are important for maintaining consciousness, movement, and executive function. It is therefore intuitive that PHE or an increase in volume in these regions could worsen damage and result in poor outcomes (Grunwald et al., 2016; Leasure et al., 2019).

It is important to highlight that the aforementioned studies have been retrospective; varied in inclusion criteria, the modality of neuroimaging, the timing of scanning, and outcome measures; and utilized heterogeneous measures of PHE. Furthermore, little is reported on the late phase of edema and how this affects outcomes after ICH. More research is needed to understand these temporal relationships in prospective studies, including a large number of patients.

Fingolimod, a sphingosine 1-phosphate receptor analog (S1 P-R), is able to cross the BBB and act directly on neurons and glial cells (Fu et al., 2014). In preclinical studies using collagenase models of ICH, treatment with fingolimod was associated with less cerebral edema and lymphocytic proliferation and decreased apoptosis (La Mantia et al., 2016). Other studies have shown that fingolimod is neuroprotective and improves behavioral outcomes (Lu et al., 2014). A phase 2 clinical study has shown that treatment was associated with reduced perihematomal edema volume (Fu et al., 2014), and this has led to testing a single-dose 0.5 mg fingolimod as potential treatment within 24 h of ictus (NCT04088630) (Wolfe, 2022).

Siponimod is a more selective S1 P-R analog but has a shorter half-life compared to fingolimod. Experimental studies of siponimod showed that it reduced cerebral edema and improved neurological recovery and survival (Bobinger et al., 2019, 2020). However, a phase 2 randomized clinical trial (NCT03338998) was terminated due to a lack of efficacy (NCT03338998, 2022).

Edaravone, which is a novel synthetic molecule with scavenging properties, is an attractive neuroprotectant. In one preclinical study, edaravone reduced brain water content, suppressed inflammation by blocking the release of NLRP3 [nucleotide binding oligomerization domain (NOD), leucine-rich repeat (LRR), and pyrin domain–containing protein 3], and improved motor and cognitive outcomes (Miao et al., 2020). Little is known of the clinical effects, and so a small trial (NCT04714177) has begun testing the effects of edaravone in hypertensive ICH patients within 48 h of onset (NCT04714177, 2021).

One study tested disufenton, a free-radical trapping agent, in patients with ICH, but there was no benefit (Lyden et al., 2007). The reasons could be reflected in the insufficient permeability of disufenton to enter through the BBB or that a single agent is not enough to neutralize the large numbers of free radicals that are generated (Lyden et al., 2007). Other agents such as pyrroloquinoline quinone and sulforaphane, which activate a protein nuclear factor (erythroid2-related factor 2), have been tested for their antioxidant properties and whether they enhance brain recovery in experimental stroke (Ma et al., 2015; Bautista et al., 2021). The clinical effects are yet to be established.

Given the inability of free radical scavengers to affect the neurological outcomes in patients with acute ischemic stroke (Serebruany, 2006; Shuaib et al., 2007) and the prominent role of pro-oxidant enzyme Nicotinamide adenine dinucleotide phosphate (NADPH) oxidase in cerebrovascular ROS generation (Serebruany, 2006; Shuaib et al., 2007; Allen and Bayraktutan, 2009), it is conceivable that the inhibition of this enzyme system or its functional subunits (p22-phox or gp91-phox) may prove efficacious. Indeed, suppression of NADPH oxidase has been shown to negate its effects in acute ischemia, hyperglycemia, or TNF-? on an in vitro model of human BBB through regulation of apoptosis, MMP-2/9 and the protein kinase C pathway (Abdullah and Bayraktutan, 2014; Rakkar et al., 2014; Shao and Bayraktutan, 2014).

Magnesium is appealing as a neuroprotectant because of its membrane-stabilizing effects, and it has been shown to increase cerebral blood flow, enhance recovery, and reduce damage from toxic free radicals (Avgerinos et al., 2019). Studies have shown that it is feasible to administer intravenous magnesium in acute stroke (Saver et al., 2004), but in one phase 3 trial, treatment within 2 h had no effect on neurological recovery or functional outcome (Saver et al., 2013). Potential explanations for these results include the small number of patients with ICH and the timing could have been too early for treatment to have any effect on edema (Saver et al., 2013).

Inhibition of excess accumulation of glutamate in acute ICH with memantine, a fast-acting, noncompetitive NMDA channel antagonist, has been tested (Lee et al., 2006). One preclinical trial of memantine reported an approximate 50% reduction in ICH volume coupled with the inhibition of MMP-9 following treatment for 14 days (Lee et al., 2006). In another study, 64 patients were allocated to either memantine (10 mg daily for a month and then increased to 20 mg daily for 2 months) or placebo within 24 h of ictus (Bakshayesh-Ebhbali and Hajinnori, 2015). At 3 months, patients treated with memantine reported better motor function compared to those receiving the placebo (Bakshayesh-Ebhbali and Hajinnori, 2015). Data also suggest that memantine may improve cognition and aphasia (Berthier et al., 2009), but this needs to be tested further.

Statins have been shown to be neuroprotective by modulating cellular pathways and controlling leukocyte proliferation, adhesion, migration, and angiogenesis (Tapia-Perez et al., 2013). Preclinical studies have shown that these effects might translate to clinical recovery through hematoma resolution and reduced cerebral edema (Jung et al., 2004; Yang et al., 2013; Chen et al., 2019). Based on these results and retrospective data showing lower mortality, the Statin for Neuroprotection in Spontaneous Intracerebral Hemorrhage trial (STATIC) is now testing the effect of Atorvastatin 20 mg in ICH-associated cerebral edema (Chen et al., 2019) (Appendix).

Other neuroprotective agents include monascin, which is a natural agonist of the PPAR. In preclinical models, monascin has been shown to reduce ICH volume and reduce cerebral edema and BBB permeability at higher doses (Wang et al., 2017; Fu et al., 2020). Further testing of this agent is awaited.

Therapeutic hypothermia and fever prevention have the potential to be therapeutic options for ICH edema (Loach and Bayraktutan, 2016; Greer et al., 2021). Preclinical studies have shown promise, and a meta-analysis concluded that hypothermia reduced edema-associated BBB disruption through the suppression of excitotoxicity, inflammation, oxidative stress, and apoptosis (Staykov et al., 2010; John and Colbourne, 2016; Loach and Bayraktutan, 2016; Melmed and Lyden, 2017). Furthermore, therapeutic hypothermia has been shown to augment the BBB-reparative capacity of therapeutic agents targeting protein kinase C-β or the NADPH oxidase enzyme system (Kadir et al., 2021). One prospective clinical study tested endovascular mild hypothermia (35 degrees celcius for 10 days) in patients with moderate volume ICH (Staykov et al., 2013). The study reported that the volume of edema remained stable throughout treatment, and, interestingly, there was no rebound increase after patients had been rewarmed to normal temperature (Staykov et al., 2013). Therapeutic hypothermia is associated with complications such as pneumonia, coagulopathy, and bradycardia, which appear to be directly related to the depth and duration of cooling (Zhao et al., 2007a; John and Colbourne, 2016). Therefore, optimizing cooling protocols might prevent such complications, but this needs to be assessed in future studies (Staykov et al., 2010; Loach and Bayraktutan, 2017).

Deferoxamine has been shown to be neuroprotective and improve outcomes in experimental ICH (Staykov et al., 2010; Selim et al., 2011). As a chelator, it reduces the amount of free iron that is available to cause cerebral edema and neurotoxicity. Deferoxamine has also been shown to have other effects: it has antioxidant properties that directly reduce neuronal damage; downregulate cytokines, particularly TNF-α; and activate the enzyme cyclooxygenase, thereby reducing brain inflammation and inhibiting toxicity caused by hemoglobin (Nakamura et al., 2004; Okauchi et al., 2009). A meta-analysis analyzing the effect of deferoxamine in preclinical models of ICH found that treatment reduced brain water by 85% and improved neurobehavioral outcomes (Okauchi et al., 2009). In addition, this analysis reported that treatment was effective if started within the first few hours of ictus (Okauchi et al., 2009).

Based on these results, a phase II clinical trial tested deferoxamine at a dose of 32 mg/kg/day (Selim et al., 2019). Participants were enrolled within 24 h of ICH, and treatment was given for 3 consecutive days (Selim et al., 2019). The results showed that deferoxamine was safe, but there was no difference in the primary outcome of death or dependency at day 90 between deferoxamine compared to placebo (34% vs. 33%) (Selim et al., 2019). A post-hoc analysis suggested a treatment effect at 6 months, indicating that the benefits with deferoxamine might be observed after a longer period but there was no effect on cognition, mood, or PHE (Wei et al., 2021). The results do not support testing of deferoxamine on its own in a future study, but combining it with another intervention could be an option (Wei et al., 2021).

Cell therapy with endothelial progenitor cells (EPCs), bone marrow–derived circulating stem cells, may induce therapeutic angiogenesis and potentially improve outcomes after ICH. Translational and clinical studies indicate that EPCs may potentiate endothelial cell regeneration, a prerequisite for effective BBB repair and neovascularization (Bayraktutan, 2019; Kadir et al., 2022a,b). Indeed, EPC levels are associated with good neurological outcomes and reduced hematoma volume in ICH, implying the potential as a novel therapeutic option for patients (Pías-Peleteiro et al., 2016).

Surgery for ICH has theoretical benefits for the management of edema: it reduces mass effect, lowers intracranial pressure (ICP), and removes inflammatory and neurotoxic blood products. Open craniotomy is the most studied approach, but two well-powered randomized controlled trials did not show a benefit of surgery compared to medical management alone (Mendelow et al., 2005, 2013). A pooled analysis of the two trials with results of previous studies suggested that early surgery irrespective of the technique was associated with a survival benefit in a subgroup of patients (e.g., those who deteriorate neurologically from hematoma expansion or superficial, lobar ICH with no intraventricular extension) (Gregson et al., 2012). It is important to highlight that the two trials were based on clinical equipoise, i.e. patients thought to benefit from surgery at the outset were excluded (Mendelow et al., 2005, 2013). Moreover, the protocols allowed cross-over of patients to craniotomy if the patient deteriorated (Mendelow et al., 2005, 2013). As a result, the actual benefit of delayed surgery in ICH is unclear and could have been underestimated.

Preclinical studies have reported that early minimally invasive surgery is associated with reduced extracellular glutamate, perihematomal endothelin-1 levels, BBB permeability, and brain water content (Wu et al., 2013; Wang L. et al., 2015). In a small study, patients with putaminal hematomas who underwent surgery had significantly reduced cerebral edema on follow-up scans (Okuda et al., 2006). However, a decrease in edema has been shown to depend on the amount of hematoma that is evacuated (Horowitz et al., 2021). Another analysis compared the effect of hematoma drainage by burr hole craniectomy on severity of cerebral edema (Zuo et al., 2009). The results showed that at 3 weeks, surgery significantly reduced levels of TNF-α and products released by the coagulation cascade (Zuo et al., 2009). Although hematoma evacuation is suggested to reduce perihematomal edema, it is unclear whether it affects outcomes. One study analyzed the relationships between the ratio of residual hematoma volume to edema, inflammation, and prognosis (Zuo et al., 2009). At 1 year, patients who had complete evacuation of hematoma had better functional outcomes (Zuo et al., 2009). A post-hoc analysis of a phase 3 trial reported a 35% reduction in perihematomal edema, but there was no improvement in functional outcomes (Hanley et al., 2019).

One factor that can increase edema is intraventricular hemorrhage (IVH) because there is an imbalance in the absorption of cerebrospinal fluid. This often leads to acute hydrocephalus, which, in turn, worsens edema by increasing ICP, reduces brain perfusion, and induces ventricular wall stretch, white matter injury, and inflammation (Holste et al., 2022). Therefore, placement of an external ventricular drain (EVD) could reduce these effects. Observational data suggest that placement of EVD in IVH is associated with reduced mortality, but the effect on long-term outcomes is unclear (Lovasik et al., 2016). It is also important to highlight that placement of EVD has its challenges including blockage, malpositioning, overdrainage, the need for prolonged drainage, and infection (Jamjoom et al., 2018). To date, there have been no randomized trials of EVD in ICH, therefore research into the potential benefits is needed.

Decompressive hemicraniectomy (DHC) refers to the removal of a window of bone with wide durotomy or dural expansion to allow for normalization of ICP and mass effect. DHC is routinely performed in malignant middle cerebral artery (MCA) infarction, but it is less established for the treatment of spontaneous ICH (Yao et al., 2018). The effect of DHC for edema after ICH is unclear, and the indication must be balanced against the complications, including hematoma expansion (~25% of cases) and an additional procedure, that is, cranioplasty (Gopalakrishnan et al., 2018). One randomized trial is evaluating whether DHC with the best medical treatment improves outcomes vs. best medical treatment alone (NCT02258919, 2023) (Appendix).

As of yet, there is no established treatment for prevention or reducing PHE, but information on which ones show promise and those that may not be effective is emerging. Some drugs used in other medical conditions have properties that could be useful in treating cerebral edema and offer the possibility of being repurposed.

Lowering BP is recommended after acute ICH, but its routine use and for secondary prevention means that specific testing for effects on PHE might necessitate a comparison of intensity (standard vs. guideline) or agent. Osmotic agents have biologically plausible mechanisms that might be effective for treating PHE, but more work is needed on the optimal dose, treatment duration, safety, and efficacy. Because hematoma volume is strongly linked with cerebral edema, surgery may still have a role, but more research is needed on the timing and technique to identify which patients benefit.

The involvement of AQP4 channels in the pathogenesis of cytotoxic edema is clear, but more work is needed to develop effective agents. The SuR1-TRPM4 channel seems to be a key target, and glyburide as a potential inhibitor is promising. One concern is that treatment might increase the risk of hypoglycemia, although this was not observed in a small phase 2 trial testing the effects of glyburide in ischemic stroke (Sheth et al., 2016).

Agents that are neuroprotective and remove toxic free radicals are promising as are treatments that attenuate inflammation and infiltration of leucocytes into the perihematomal region after ICH. With careful selection, it is possible that some of these could translate from preclinical to clinical studies. This review also highlights that a better understanding of the role of inflammation and the immune system after stroke is needed as the potential for modulation could have profound implications for clinical management.

As highlighted, the mechanisms of PHE are complex with the magnitude time-dependent, and this should be taken into consideration when designing future clinical studies. For example, a trial could test reducing hemorrhage expansion first and, when the bleeding has stopped, then assess evacuation followed by an anti-inflammatory or neuroprotective agent. In the acute phase, it may be that osmotic agents complement BP lowering, hemostatic therapy, or surgery for ICH. Such an innovation would warrant streamlining of trial protocols, careful patient selection, and incorporation of specific outcomes to test response and recovery. This work will need close collaboration between clinicians, researchers, pharmaceutical companies, and industry.

The present review also highlights that there is variation in the measurement of PHE and the modality of brain imaging, therefore more research is needed to identify a standard to assess change or severity. Evidence suggests that brain repair and recovery can continue for months after ICH (Selim et al., 2019), therefore testing the long-term effects of potential treatments for cerebral edema is warranted.

KK: Conceptualization, Writing—original draft, Data curation, Methodology, Resources. PC: Writing—original draft, Data curation, Resources. TN: Writing—review and editing. CT: Data curation, Writing—original draft, Resources. SC: Data curation, Writing—original draft, Resources. JA: Writing—review and editing. ZL: Writing—original draft, Resources. MH: Writing—original draft. MK: Writing—review and editing. TE: Writing—review and editing. CR: Writing—review and editing. MM: Writing—review and editing. JD: Writing—review and editing. UB: Writing—original draft. DW: Writing—review and editing. NS: Writing—review and editing. PB: Writing—review and editing.