Cerebral swelling is a common feature of CM in adults and children in radiological studies [8–12]. The degree of swelling varies, and the radiological features usually suggest that it results from excess intracranial blood volume. Intracranial pressures are often elevated in children with CM but not in adults [27, 28]. Not all cerebral swelling seen radiologically or macroscopically at autopsy results from parenchymal cerebral oedema. Radiological evidence of frank oedema is less common although it is well described, particularly in fatal cases near to death. There are two main types of oedema recognized in the brain . Vasogenic oedema is characterized by leakage of water into the perivascular and parenchymal extracellular compartments resulting from BBB breakdown. Cytotoxic oedema represents cellular swelling resulting from the breakdown of the normal membrane controls of fluid and ion transport that maintain osmotic homeostasis in neurones and other cells such as astroglia .
In this autopsy study, histological examination showed that microscopic oedema was a common finding in the post-mortem brains of fatal Vietnamese adults who died of severe malaria cases, but that brain weights were not significantly increased and that evidence for microscopic cerebral oedema was not associated with coma pre-mortem.
There are many reasons for the brain to be swollen in severe malaria. Seizures are often multiple and protracted (particularly in children) and there may be metabolic alterations because of liver dysfunction or renal failure (predominantly in adults). In acute cerebral malaria there is marked intravascular congestion of parasitized erythrocytes and uninfected erythrocytes resulting from sequestration, which increases the intravascular blood volume and therefore brain volume without altering overall cerebral blood flow. Apart from the disease process itself, treatment may result in excess fluid administration, or mechanical ventilation may undercompensate for the hypocapnia that accompanies metabolic acidosis and cause cerebral vasoconstriction.
Cerebral oedema can be variable and focal, around defined larger areas of white matter damage, so may not make a significant contribution to the overall brain water content ('volume' or 'weight') but might still have a negative impact on neuronal function. The host reaction to PRBC sequestration may differ in young patients with less multiorgan disease, perhaps because of relative functional immaturity of the BBB making oedema more likely in younger patients. In vivo studies of BBB function through examination of CSF protein partitioning support a greater degree of BBB breakdown in African children  compared to Vietnamese adults  although the degree of permeability and the intracranial pressure increases are much less than in pyogenic meningitis. A recent paper has examined the development of cerebral oedema in murine model of experimental cerebral malaria. This model has been known for some time to demonstrate increased vascular permeability associated with fatal outcome in Plasmodium berghei-infected mice. The results of this study indicated a strong association between cerebral oedema and experimental cerebral malaria in mice , which was not found in adult human fatal malaria cases.
Because of these remaining uncertainties as to the role of brain oedema in the pathogenesis of coma in human CM, the prevalence and patterns of cerebral oedema in post-mortem brain from adult fatal malaria cases was examined, and related to factors which could influence the formation of oedema, including systemic complications of disease such as renal failure, immunohistochemical evidence of vascular integrity at the blood-brain barrier, and the expression of the water transport channel Aquaporin 4. Several distinct histological forms of oedema were identified and found to co-exist in the same tissue section. The categorization of four different patterns of histological oedema used in this study was based initially on the accepted neuropathological differences between cytotoxic oedema and vasogenic oedema . Examination of this series showed recognizable separate patterns within these, such as bubbly parenchymal oedema rather than pericellular cytotoxic oedema, or deposition of granular proteinaceous material rather than just perivascular clearing in vasogenic oedema. These have not been previously described and as such the classification of the neuropathological features of oedema described here contains novel classes, which may be inter-related.
The prevalence of loss of vascular integrity was heterogeneous within patient groups. Using both post-mortem and in vivo studies of CSF previous studies had demonstrated disruption of BBB function, but in a mild and reversible form that is not as marked as other inflammatory and infective encephalopathies . In the current study, there were no significant differences in the prevalence of oedema or haemorrhages, or loss of vascular integrity, between CM and fatal malaria cases without neurological complications. Therefore there was not a simple relationship between BBB disruption, fibrinogen leakage, development of oedema and subsequent coma.
However, there were differences in oedema formation and patterns of fibrinogen leakage between different brain regions. In general, there was a greater degree of oedema in brainstem compared with cortex and to a lesser extent with diencephalon. For example, evidence of oedema between axonal fibre tracts could be observed in all brainstem sections but was observed in less than 30% of cortical sections from severe malaria cases. Extravasation of proteinaceous material into the perivascular Virchow-Robbins space in the cortex was more common in patients with longer survival times after admission, suggesting that these changes need time to develop in order to be observed pathologically. However, these correlations were not observed in the brainstem despite a high incidence of perivascular protein leakage.
Immunohistochemical analysis showed that fibrinogen leakage in the cortex was frequently diffuse around small vessels whereas the brainstem displayed fibrinogen leakage around both small and larger vessels in a cell-associated or diffuse pattern. Correlations between patterns of fibrinogen leakage and markers of organ dysfunction were observed in the brainstem but not in the cortex or diencephalon. These results can be interpreted in several ways. The brainstem may be more susceptible to vascular damage or parenchymal cell injury in severe malaria cases that have a fatal outcome, or perhaps different mechanisms of cell damage could be in operation resulting from differences in tissue architecture. The lack of difference between CM and non-CM cases suggests the observed loss of vascular integrity is unlikely to be the cause of cerebral symptoms in CM per se.
The VEGF pathway has been demonstrated to increase BBB permeability, so the correlation between evidence of VEGF signalling and evidence of BBB leakage including fibrinogen leakage or prevalence of histological types of oedema was examined. However, there were no correlations with pKDR immunolabelling. Despite this, involvement of the VEGF pathway in the modulation of vascular integrity during severe malaria cannot be completely ruled out, since the peak of VEGF signalling may have occurred before treatment or death, and the parasite itself may modulate VEGF signalling in severe malaria . Other potential vasomodulatory pathways have been examined in the severe malaria, such as tumour necrosis factor, angiopoetin [35, 36], erythropoietin and nitric oxide. Inflammatory processes within the brain may also synergize VEGF expression, given that CD-8 positive T-cells can induce BBB leakage due to neuronal VEGF expression in neuroinflammatory conditions .
In contrast with VEGF, a statistically nonsignificant trend was identified for AQP4 staining to be higher in the brainstem of CM cases compared with non-CM cases (P = .02). Brain injury can variably up- or down-regulate AQP4 expression depending on the disease studied or the experimental model used, in the face of retained GFAP expression . AQP4 induction may have different roles at different time points in disease. Since AQP4 controls water fluxes into and out of the brain parenchyma it remains difficult to determine whether the increased expression contributes to oedema formation or is a sign of neuroprotective attempts at resolution. In this study, there was no correlation between AQP4 expression levels and the loss of vascular integrity in the severe malaria sections. However, previous studies have shown that enhanced AQP4 immunoreactivity can be detected around ischaemic foci even after brain oedema had resolved . One possible scenario in severe malaria is that AQP4 does not prevent opening of the BBB or vasogenic oedema formation but, rather, promotes water clearance from the brain, limiting the development of cytotoxic oedema . This would normally develop following extensive microvascular obstruction, similar to that directly observed in retina and rectal mucosa of CM patients [41, 42]. Gene knockout studies in mice suggest multiple roles for AQP4 in regulation of water transport, astrocyte signalling and direct modulation of neuronal excitability [43, 44]. In this study, AQP4 labelled processes were found around neuronal somata. Increased seizure duration and slowed potassium kinetics occur in mice lacking AQP4 channels . This has been attributed to subcellular co-localisation of AQP4 with the inwardly rectifying potassium channel Kir 4.1, suggesting that AQP4 may participate in the coupled influx of water and K+ that occurs after neural activity. The frequent occurrence of seizures as part of the clinical spectrum of CM may influence the pattern of AQP-4 expression.
This study was designed to determine any association between brain oedema and the neurological complications of severe malaria, in order to validate approaches to potential trials of neuroprotective adjuvant treatment. Autopsy based studies clearly have disadvantages in following the temporal sequence of cerebral swelling in CM. However planned studies differentiating between cytotoxic and vasogenic oedema on the basis of diffusion weighted functional MRI imaging should allow some comparison of these post-mortem findings with malaria patients in vivo . Functional MRI imaging offers a promising way to study the effects of oedema on metabolic and pathological changes in both human CM and relevant animal models of the disease 
Recent treatment studies in both African children  and Indian adults  have confirmed that although cerebral swelling and associated increases in intracranial pressure often occur in paediatric cases of CM, treatment with mannitol, which reverses vasogenic oedema, is ineffective or harmful. Steroids, which also reduce oedema as well as inflammation, are also ineffective in adult CM . Therefore, the processes by which malaria cause brain swelling and the relationship of this swelling to extravascular oedema are not straightforward, and may not offer the potential for adjuvant neuroprotective therapy.