Although the complex mechanism of PHE growth is not yet completely clear, preclinical evidence suggests that different pathophysiologic mechanisms dominate the development of PHE at different stages of ICH ( Figure 3 ) (4–6, 25). The essential phenomenon underlying PHE formation is an imbalance of the perivascular fluid interchange based on Starling’s principle (26). Specifically, the formation of an osmotic gradient and the elevation of capillary hydrostatic forces drive the flow of fluids that result in the development of PHE. Starling’s classic formula is J _V = K _H(P _C − P _I) − K _O(π_C − π_I), where J _V is the net transendothelial fluid transfer in brain capillaries; K _O and K _H are the filtration coefficients of oncotic conductivity and hydraulic conductivity, respectively; π _C − π _I is the difference in oncotic pressure between the capillaries and interstitial tissue; and P _C – P _I is the difference in hydraulic pressure between the capillaries and interstitial tissue.

Both cytotoxic edema and vasogenic edema play particularly significant roles in PHE formation. Cytotoxic edema dominates the initial stage of PHE and is a premorbid precursor to extracellular ionic edema resulting from the dysfunction or abnormal activation of ion pumps in endothelial cells and astrocytes. Perihematomal glutamate deposition may contribute to cytotoxic edema (27). The extracellular concentration of glutamate in patients with stroke or traumatic brain injury can be 20 times higher than that in healthy individuals (28, 29). The opening of ion channels increases the movement of water from the extracellular to intracellular space, causing cell swelling and even cell death (30, 31). The essential mechanism of cytotoxic edema is the variation in the brain water distribution, which does not induce true tissue space swelling; however, the transendothelial osmotic gradients derived from cytotoxic edema provide the driving force for ionic edema. Astrocyte swelling is a typical manifestation of cytotoxic edema. When cytotoxic edema occurs, water gains access to the central nervous system through the AQP-4 expressed in astrocytic foot processes (32). Upregulated AQP-4 expression has been identified in patients with ischemic stroke and facilitates ionic edema formation (33). However, how AQP-4 affects ion transcellular transport remains unclear.

Vasogenic edema dominates the second stage of PHE formation, which is characterized by BBB dysfunction caused by a series of neuroinflammatory responses associated with the mechanical destruction of ICH, thrombin activation, and toxic effects of erythrocyte lysis (4–6). Vasogenic edema is a consequence of multifactorial actions. In the immune response associated with neuroinflammation, the disruption of tight junctions between vascular endothelial cells increases vascular permeability via inflammatory cell chemotaxis, cytokine and chemokine release, and upregulation of vascular endothelial growth factor (VEGF) and matrix metalloproteinase-9 (MMP-9) (4, 5, 26). The endothelial cell swelling and cell membrane breakdown caused by cytotoxic edema may also increase the permeability of the BBB (34). After BBB injury, both the filtration coefficients of oncotic conductivity and the hydraulic conductivity rise (26), the water and macromolecular substances can more easily pass through the cell membrane and enter the interstitial tissue of the brain, causing vasogenic edema. The intracranial pressure, blood pressure, and concentration of intravascular osmotically active molecules influence the relevant hydraulic pressure and oncotic pressure, thereby affecting the formation of PHE (26). BBB opening is associated with rapid activation of the complement cascade. Complement fragments (e.g., C3a, C5a, and others) amplify the inflammatory response in a positive feedback loop to disrupt the BBB (35), producing anaphylatoxins and membrane attack complexes that lyse erythrocytes and thus promote the formation of iron-induced PHE (36).

In the first few hours after ICH onset, during which the coagulation cascade is activated, the blood clot retraction, cell death, and brain atrophy induced by destruction of the hematoma produces a relatively large perihematomal space, leading to a reduction in the perihematomal hydrostatic pressure (37). The serum protein that is extruded secondary to the blood clot retraction leads to an increase in the interstitial oncotic pressure (38). Together, these changes induce the initial transport of water into the brain tissue, leading to edema. An important point to note is that the ionic edema driven by the cytotoxic edema dominates the first stage of PHE. The development of cytotoxic edema reportedly involves aberrant regulation of ion transport channels expressed on vascular endothelial cells (e.g., Na-K-Cl cotransporter 1 (NKCC1), which plays a role in the brain edema associated with ischemic stroke) and the sulfonylurea receptor 1–transient receptor potential cation channel subfamily M member 4 (SUR1-TRPM4) (39).

In the early stage of ICH, after transition from the closed to open state of the NKCC1 secondary to perihematomal glutamate deposition (28), Na⁺ is transported across the membrane to the interstitial tissue; this transport is driven by the transendothelial forces produced by the cytotoxic edema. Ionic edema forms as the Cl⁻ and water follow the movement of Na⁺ to maintain electrical and osmotic neutrality (40). When adenosine triphosphate depletion occurs, the NKCC1 closes while the SUR1-TRPM4 opens, allowing the Na⁺ and water to be transferred to the interstitial tissue along their gradient, resulting in edema (39). Nevertheless, the associated molecular pathways of ionic edema remain poorly understood, and the involvement of cytotoxic edema in PHE is not well defined.

Within 2 days after ICH onset, the vasogenic edema induced by the inflammatory immune response dominates the second stage of PHE (4–6). Numerous molecules are involved in this response, and four main pathways have been recognized. First, the thrombin and mechanical destruction of the hematoma activate Toll-like receptor 4 and the nuclear factor κB (NF-κB) pathway. NF-κB activates and regulates the transcription of cytokines, chemokines, and MMPs, leading to BBB dysfunction (41). The expression of Toll-like receptor 4 begins at 6 h after ICH onset and persists for almost 7 days, also triggering microglial activation (42, 43). Second, the activated thrombin induces the expression of chemokines and adhesion molecules, promoting the recruitment and infiltration of inflammatory cells (e.g., neutrophils, macrophages, and lymphocytes) to perihematomal sites (44). These recruited inflammatory cells release cytokines, reactive oxygen species, tumor necrosis factor-α, and MMPs, leading to BBB injury (45). Chemotaxis of neutrophils and polymorphonuclear leukocytes begins shortly after ICH onset and peaks at 3 days (46). Third, thrombin can further activate astrocytes and microglia via proteinase-activated receptor 4. The hyperactivation of microglia may exaggerate neuroinflammation through the secretion of reactive oxygen species, tumor necrosis factor-α, and cytokines (4, 5). Because microglial activation peaks at 3 days and significantly decreases 1 week after ICH (47), the SBI induced by these activated microglia is still maintained despite the fact that the leukocyte infiltration gradually resolves after 2 to 3 days (4). Notably, M2 microglia promote endogenous clearance of the hematoma following ICH (48). However, in the acute and subacute phases, the M2-dominant microglia quickly switch to M1-phenotype microglia (49); these M1 microglia excessively release destructive proinflammatory mediators and neurotoxic substances, leading to BBB dysfunction, PHE, and neurologic dysfunction (50). Thus, conversion of M1 to M2 microglia may be a potential treatment modality for ICH-induced SBI. Fourth, the activated complement cascade increases the production of anaphylatoxins and chemokines, resulting in increased permeability of the BBB (35).

Although erythrocyte lysis is initiated within 24 h after ICH (4), erythrocyte lysis with resultant hemoglobin and iron-related toxicity still dominates the PHE process 3 days after ICH onset. Specifically, the erythrocytes are dissolved to hemoglobin by complement-produced membrane attack complexes, and the hemoglobin is then oxidized into methemoglobin, which rapidly liberates its heme. The heme is then degraded into free iron via heme oxygenase enzymes (36). An experimental rat model showed that iron deposition occurs within 24 h after ICH, peaks after 7 days, and is maintained at a high concentration for at least 2 weeks (51). The free iron also stimulates the production of reactive oxygen species and MMP-9, promoting an inflammatory reaction and BBB dysfunction (45). The deposition of hemosiderin upregulates AQP-4, which exacerbates the brain edema and peaks at 3 to 7 days (5, 52). The hemoglobin and heme can also directly activate Toll-like receptor 4, microglia, and the NF-κB pathway to further promote the inflammatory reaction (53, 54). Consequently, the third stage of PHE is also arguably a delayed stage of vasogenic edema induced by erythrocyte lysis.

Numerous studies have demonstrated that the initial hematoma volume determines PHE formation (9, 10, 12, 65), which shows good agreement with the aforementioned mechanism in which a greater ICH volume is associated with stronger thrombin cascades, erythrocytes lysis, and ICH-related toxicity. Several studies have revealed that a higher percentage of surgical hematoma removal results in slower PHE growth (24, 66). However, Sprügel et al. (20) indicated that the surface of the hematoma, not the initial hematoma volume, is the primary driver of PHE growth because a smaller hematoma has a larger relative surface area, contributing to a higher PHE volume per unit of the hematoma’s surface area. The evidence that irregular ICH and relatively minor ICH (<30 ml) generate a higher relative PHE volume was verified, supporting the surface-driven hypothesis. Notably, however, irregular and relatively minor ICH has a higher relative, not absolute, PHE volume. A recent study showed that certain CT imaging signs, such as the blend sign, black hole sign, and island sign, are capable of predicting hematoma expansion (67) and are associated with PHE growth in the acute phase of ICH (68). However, there is currently no evidence supporting an association between hematoma expansion and PHE formation. Rodriguez-Luna et al. (12) found that patients with spot signs on baseline CTA had a larger absolute PHE volume. Nevertheless, using the absolute PHE to predict hematoma expansion would be inappropriate because the PHE strongly depends on the initial hematoma volume. There is great controversy regarding the severity of PHE in different ICH locations. Sprügel et al. (20) found that lobar ICH had a larger initial PHE volume and higher early PHE growth rate. However, there was no significant difference in the peak PHE volume between deep and lobar ICH after adjusting for the hematoma volume (20). McCarron et al. (69) found that PHE was not affected by the location of ICH within 24 h after onset. In contrast, Grunwald et al. (11) found that the growth rate of lobar PHE within 24 h of ICH was significantly higher than that of deep ICH and was associated with the 90-day mortality rate. However, there was no significant difference in the growth rate of PHE within 72 h of onset between lobar and deep ICH. Cerebral amyloidosis was found to be a common cause of lobar ICH, which has localized anticoagulant and thrombolytic properties (70). However, we found no evidence indicating that lobar PHE is significantly smaller than deep PHE. We speculate that different shapes and growth patterns of PHE exist in different locations, and these differences are probably due to the different morphologies of the hematoma and the heterogeneity of the targeted population with diverse characteristics of ICH.

In addition to imaging features, the patient’s baseline neurological status (e.g., as measured by the National Institutes of Health Stroke Scale score, Glasgow Coma Scale score, and other indexes) is also significantly associated with PHE progression (10). Advanced age is an independent risk factor for PHE (10, 61). However, Peng et al. (21) indicated that younger patients are more likely to develop delayed PHE formation, which may be due to age-related differences in brain atrophy. Whether sex influences PHE continues to be debated. Wagner et al. (71) found that the PHE volume in women with supratentorial ICH is lower than that in men which may be associated with the higher levels of estrogen in women, enabling alleviation of iron-induced PHE. However, other studies have produced different or even contrary conclusions (16, 65). Because poorly controlled hypertension and unstable blood pressure at admission are assumed to be risk factors for hematoma expansion (72, 73), the role of blood pressure in PHE is also attracting interest. The Intensive Blood Pressure Reduction in Acute Cerebral Haemorrhage Trial (INTERACT) showed that a history of hypertension was positively correlated with the relative growth of PHE, whereas lower systolic blood pressure at admission was positively associated with the absolute growth of PHE (65). In the ICH Acutely Decreasing Arterial Pressure Trial (ICH ADAPT) performed by McCourt et al. (74), aggressive antihypertensive treatment (diastolic blood pressure of <150 mmHg) was found to affect neither the perihematomal cerebral blood flow nor PHE progression. A clinical trial of hypothermia administration for PHE revealed that a higher number of hypertension attacks at admission was associated with a larger initial PHE volume (61). Moreover, the Antihypertensive Treatment of Acute Cerebral Hemorrhage-2 (ATACH-2) trial also demonstrated that intensive antihypertensive therapy (target systolic blood pressure of 110–139 mmHg within 2 h) effectively reduced the relative expansion rate of PHE within 24 h after onset (62). In general, intensive antihypertensive therapy has been demonstrated to be safe in the treatment of PHE. However, whether the control of blood pressure mediates the PHE by modulating hematoma expansion is unclear. Previous studies have shown that a shorter interval between onset and the initial CT scan is associated with a higher risk of hematoma expansion and more rapid relative PHE growth (65, 75). However, Rodriguez-Luna et al. (12) indicated that the ICH onset time did not affect the relative PHE volume within 6 h of onset. Because obtaining a precise onset time is significant for determining the optimal treatment of ICH, whether the onset time affects the PHE needs to be further investigated. Additionally, the imaging features of ICH and PHE may help to predict the onset time.

Several laboratory parameters that have been confirmed to affect hematoma expansion, such as hyperglycemia, a high MMP-9 level, and a high white blood cell count, may also be positively associated with PHE progression (10, 61, 76, 77). Gusdon et al. (78) found that the ratio of neutrophils to lymphocytes is effective in predicting PHE growth. Because MMP-3, MMP-9, VEGF, and angiopoietin-1 are all related to vascular function, they might be predictive of vasogenic edema (79). A high RBC count and high hematocrit at admission are associated with delayed peak PHE (16). This may be relevant because a high RBC count and high hematocrit are indicative of higher RBC degradation, which has been identified as an essential factor for promoting PHE. A high platelet count promotes increases in the VEGF level and capillary permeability, thereby exacerbating PHE (16). A prolonged partial thromboplastin time is significantly associated with PHE growth (16, 61, 80). This could be due to a consumptive coagulopathy resulting from significant release of coagulation factors after ICH, manifesting as platelet dysfunction and prolonging the partial thromboplastin time (80, 81). Some studies have shown that hyperglycemia is associated with earlier PHE progression (10, 82), which might be due to the fact that hyperglycemia promotes an oxidative stress response with consequent BBB dysfunction (83). However, hyperglycemia can merely be a stress response to ICH instead of a contributing factor for PHE. Feng et al. (84) found that hyperglycemia did not significantly affect PHE after adjusting for the initial ICH volume.

Apolipoprotein E (Apo­E) has been considered an independent risk factor for lobar ICH (85). Apo-E plays an essential role in maintaining normal lipid homeostasis in the central nervous system (86) and mitochondrial resistance to oxidative stress (87). James et al. (88) found that APOE-ϵ4 positivity was associated with a larger PHE volume. However, McCarron et al. (89) demonstrated that APOE-ϵ4 positivity was not associated with PHE after adjusting for race, age, and type of bleeding. These two studies reached different conclusions, which might be related to the selection of different time windows for PHE observation. Further studies are still needed to understand the role of Apo­E in PHE progression.

The formation of the perihematomal osmotic gradient, which is driven by cytotoxic edema, dominates PHE in the ultra-early phase of ICH (25, 38). Compared with mannitol, the currently available dehydrating drug that is commonly used to decrease the intracranial pressure, continuous infusion of hypertonic saline has been identified as a safe method for controlling PHE progression in the early phase of ICH, and it does not seem to affect the BBB (100, 101). Given the putative advantages of hypertonic saline in improving cerebral perfusion, Cook et al. (102) reported that hypertonic saline possesses a better capacity for controlling PHE and the corresponding intracranial hypertension than mannitol. However, in a recent multicenter randomized controlled trial, Roquilly et al. (103) found that continuous infusion of 20% hypertonic saline solution did not improve the neurological outcome 6 months after onset among patients with moderate to severe traumatic brain injury. As previously described, during the stage of cytotoxic edema, the SUR1-TRPM4 channel was confirmed to be upregulated, promoting ionic edema (5, 104). Jiang et al. (105) established a model of autologous blood-induced ICH and found that glibenclamide (a SUR1 inhibitor) effectively reduced the PHE volume, which was associated with cognitive deficit improvement. However, another study involving a model of collagenase-induced ICH showed that glibenclamide neither aggravated nor ameliorated the PHE volume or neurological dysfunction (106). Sheth et al. (107) conducted a double-blind, randomized controlled trial of patients with cerebral hemispheric infarcts and found that glibenclamide therapy significantly reduced the midline shift and MMP-9 level compared with the control group, revealing the potential role of glibenclamide in alleviating PHE after stroke. Additional clinical trials are needed to investigate preclinical strategies for cytotoxic edema.

The thrombin cascades, inflammatory response, and BBB dysfunction have been confirmed to exert essential functions in PHE formation (4). Thus, treatments targeting critical molecules in the formation of vasogenic edema (such as VEGF, MMPs, and AQPs) may be promising (5, 108). In one study, the anti-inflammatory drug fingolimod alleviated the progression of PHE and improved the functional independence of patients with ICH at 90 days (109). However, patients with a sizeable initial hematoma (>30 ml) were not included. Given that PHE is strongly dependent on the primary hematoma, whether fingolimod can benefit critically ill patients with ICH remains to be explored. Statins are HMG-CoA reductase inhibitors that exert their neuroprotective effects by anti-inflammatory actions and facilitation of neo-angiogenesis (110). Statins have also been found to reduce the absolute or relative PHE volume (111), and most relevant studies have shown that statin use does not increase the risk of ICH recurrence (112). However, the effect of statins on the growth rate of PHE has not been demonstrated. Celecoxib is a selective cyclo-oxygenase 2 receptor inhibitor that attenuates the inflammatory reaction and edema by inhibiting the generation of prostaglandins (113). A multicenter randomized controlled trial confirmed the efficacy of using celecoxib to reduce the expansion rate of PHE (114). However, because the time to initial CT was longer in the celecoxib group than in the control group in that study, the primary outcome was defined as a ≥20% change in PHE from onset to an average of 1 week, which may be inappropriate because a longer time to initial CT may represent a steady state for PHE (10, 98). Antiadrenergic drugs such as β-blockers and α2-agonists have also been used to manage hypertension in patients with ICH. A retrospective analysis of a prospective cohort of patients with cerebral hemorrhage (CHANT trial) showed that the administration of antiadrenergic drugs effectively reduced PHE within 72 h after onset (115), suggesting that a reduction of central/peripheral sympathetic activity attenuates neuroinflammation and thereby alleviates the PHE. Notably, the reduction of PHE might not have been due to the antihypertensive actions of these antiadrenergic drugs because other kinds of blood pressure-lowering drugs did not result in the same degree of PHE reduction. The transcription factor peroxisome proliferator-activated receptor gamma (PPAR-γ) plays a significant role in modulating the biomarkers of oxidative stress and inflammation (116). One study showed that the PPAR-γ agonist rosiglitazone significantly reduced the expression of proinflammatory genes such as tumor necrosis factor-α, interleukin-β, and MMP-9 in a rat model of ICH and consequently attenuated the SBI (117). However, there is a scarcity of clinical trial data regarding the use of PPAR-γ agonists in patients with ICH. The iron chelator deferoxamine is a potential candidate for ICH treatment, and its effectiveness in alleviating PHE has been confirmed in experimental models of ICH (118, 119). A meta-analysis of the efficacy of deferoxamine in an experimental ICH model showed that deferoxamine reduced the brain water content by 85.7%, although the effect lasted for only 24 h after onset (120). However, it is discouraging to note that a double-blind, randomized controlled clinical trial showed no association between administration of deferoxamine mesylate and better neurological outcomes in patients with ICH (121). Indeed, because deferoxamine is characterized by a small effect size, it would be better to enroll a considerably high number of patients to verify the drug’s utility and validity when conducting studies targeting a small effect size. Li et al. recently performed an unbiased genome-wide transcript sequencing study for surgical removal of perihematomal brain tissue in patients with ICH and identified abundant expression of formyl peptide receptor 1 (FPR1), which promotes neuroinflammatory reactions. Under the screening of a computer-aided drug design system, the research group further selected an FPR1 inhibitor (T-0080) that can cross the BBB and successfully reduced the PHE by about 35% in experimental ICH models to improve the neurological status (122). This FPR1 inhibitor may be a promising candidate for ICH therapy. Antidiuretic hormone maintains the brain water content by regulating the permeability of capillaries. Conivaptan is an antidiuretic hormone receptor antagonist that was confirmed to reduce brain edema and repair the BBB function in an experimental ICH model (123). The effects of conivaptan on the treatment of PHE might be correlated with a reduction in the expression of AQP-4 (124). A recent phase I clinical trial verified the safety of conivaptan in the treatment of PHE (125). A phase II clinical trial of antidiuretic hormone receptor antagonists is urgently needed to further explore their efficacy on SBI in patients with ICH.

The MISTIE II trial showed that minimally invasive surgery combined with tissue-type plasminogen activator effectively reduces the PHE volume (126). In the MISTIE III trial, the patients in the surgery group had improved neurological outcomes at 1 year when no more than 15 ml of hematoma remained at the end of the treatment (127). Similarly, a recent study showed that minimally invasive endoscopic surgery for ICH evacuation alleviated the postoperative PHE progression. A higher percentage of hematoma removal results in slower PHE growth (24). Although hematoma evacuation surgery has been shown to be a promising treatment for PHE, whether the patient’s prognosis can be significantly improved remains unclear. Fung et al. (23) reported that patients who underwent simple decompression surgery (without hematoma removal) developed more severe PHE than patients in the control group. However, there was no significant difference in the 90-day neurological function between the decompressive craniectomy group and the control group (23). Perhaps when the surgical indications for ICH become more detailed and standardized, the benefits of surgical interventions in reducing the complications of SBI (e.g., PHE) will gradually emerge.

Supportive treatments, including hypothermia therapy (61, 128), intensive antihypertensive therapy (62), and hyperbaric oxygen therapy (129), have also been reported to reduce PHE growth. However, the number of existing clinical studies is small, and inevitable bias and confounding factors have limited these studies. A large‐scale prospective follow‐up study for validation is warranted.