Diabetic Ketoacidosis-Associated Stroke in Children and Youth

Diabetic ketoacidosis (DKA) is a state of severe insulin deficiency, either absolute or relative, resulting in hyperglycemia and ketonemia. Although possibly underappreciated, up to 10% of cases of intracerebral complications associated with an episode of DKA, and/or its treatment, in children and youth are due to hemorrhage or ischemic brain infarction. Systemic inflammation is present in DKA, with resultant vascular endothelial perturbation that may result in coagulopathy and increased hemorrhagic risk. Thrombotic risk during DKA is elevated by abnormalities in coagulation factors, platelet activation, blood volume and flow, and vascular reactivity. DKA-associated cerebral edema may also predispose to ischemic injury and hemorrhage, though cases of stroke without concomitant cerebral edema have been identified. We review the current literature regarding the pathogenesis of stroke during an episode of DKA in children and youth.


Introduction
Type 1 diabetes mellitus (T1DM) is a common autoimmune condition that often presents in childhood and may be complicated by episodes of diabetic ketoacidosis (DKA). DKA is a state of severe insulin deficiency, either absolute or relative, resulting in hyperglycemia, ketonemia, acidemia, and systemic inflammation. Compared with adults, episodes of DKA in children carry a higher risk of morbidity and mortality. This is predominantly attributable to intracerebral complications [1][2][3][4][5], which occur in 3-10 pediatric patients per 1000 cases of DKA [6]. The most common intracerebral complication of DKA is cerebral edema (DKA-CE), which results in the death of 21-24% of affected patients, and significant morbidity in a further 10-35% [6][7][8]. Less common, and perhaps underappreciated, is the risk of acute ischemic or hemorrhagic stroke during the acute DKA episode. It has been estimated based on case series that approximately 10% of intracerebral complications of DKA are due to hemorrhage or ischemic brain infarction [4,5]. While some cases of brain infarction may arise secondary to DKA-CE-induced herniation with resultant vessel occlusion, it has become clear that not all cases of stroke in DKA are associated with cerebral edema (CE). As the presentation of stroke associated with DKA may mimic that of CE but requires different management strategies, it is imperative for the clinician to be cognizant of this potential complication. This review will examine the etiology and pathogenesis of stroke associated with episodes of DKA in children and youth (age <20). While Type 2 diabetes mellitus is becoming more commonly recognized in the pediatric population and may present with, or be complicated by, DKA [9,10], cerebral thrombosisassociated stroke has not been reported with Type 2 diabetes mellitus. Thus, the review focuses on children with T1DM.

Pathogenesis of Ischemic Stroke
Diabetes mellitus is a known independent risk factor for ischemic stroke, conferring two times the risk of an ischemic event in adults compared to the nondiabetic population 2 Stroke Research and Treatment [11,12]. There are several characteristics of DKA that place children at higher risk of cerebral ischemia. The reported cases on DKA-associated stroke in children and youth are presented in Table 1.
Chronically, the vascular endothelium is a primary target of the abnormal glycemic metabolism in T1DM [27]. Children with T1DM may be at risk of a chronic state of inflammation and endothelial activation outside of episode of DKA. Children within 1 year of diagnosis have been reported to have biochemical evidence of inflammation, with increased levels of both serum prothrombin fragments and TNFα compared to children more than 1 year post-diagnosis and to nondiabetic controls [36]. Furthermore, this report detailed evidence of endothelial perturbation, characterized by levels of von Willebrand Factor (vWF) and tissue plasminogen activator (tPA) more than 2 standard deviations higher than control. Another study found that the endothelial cellspecific adhesion molecule, soluble endothelial leukocyte adhesion molecule (sE-Selectin), was elevated in children with T1DM compared to healthy controls, and positively correlated with serum glucose concentration [37]. Analysis of the coagulation system in adults with diabetes has also identified abnormalities in many steps of the coagulation system [38].

Abnormalities in the Coagulation Cascade.
Two case controlled studies [42,43] found an increased rate (50%) of clinically apparent deep venous thrombosis (DVT) in very young children (less than 3 years) with DKA who required femoral central venous catheter (CVC) insertion when compared with age-matched nondiabetic controls (who also underwent femoral CVC insertion). Comparatively, an incidence rate of 1.5-18.3% has been described for clinically or radiologically apparent femoral CVC-associated DVTs in the PICU population [42,44]. Although the propensity for hypercoagulability in diabetes mellitus has not been described as a specific isolated risk factor for DVT in children [42], it is clear that the procoagulant mechanisms that place children with an episode of DKA at risk of CVC-related DVT may also act to increase the risk of stroke.
The inflammatory condition seen in DKA, with endothelial perturbation, predisposes to an acquired procoagulant state [26]. While the majority of case studies of children with DKA-associated stroke have not identified consistent, generalized alterations in the coagulation system (Table 1), these studies have mainly examined the coagulation system at a single time point after identification of a neurologic abnormality. More systematic evaluation of coagulation abnormalities during an episode of DKA requires longitudinal consideration, both before and during therapeutic intervention. Indeed, longitudinal studies in children [41] and adults [35] with DKA have identified multiple coagulation abnormalities, including increased platelet aggregation, elevated levels of procoagulants, and decreased activity of anticoagulants. Coagulation factors for which abnormalities have been noted during DKA or its treatment in either children or adults have been summarized ( Table 2).
Examination of the coagulation factors of 7 adolescents on presentation with DKA and at several time points after initiation of DKA therapy demonstrated abnormalities in Protein C, Protein S, plasma homocysteine, and von Willebrand Factor (vWF) [41]. Protein C levels are initially elevated but quickly decrease to normal with DKA treatment, while Protein C activity is initially low and slowly normalizes with treatment. Adult patients with T1DM have significantly lower protein C levels than controls [45]. Their protein C levels are inversely related to glucose concentrations but exhibit no relationship with glycosylated hemoglobin A (HbA1c) levels. This latter finding suggests that acute, rather than chronic, variations in blood glucose may determine the response of Protein C, which may itself explain the normal values found in most patients with DKA-associated stroke ( Table 1).
Plasma homocysteine is an important factor in atherosclerosis and thrombosis [46] and also decreases protein C activation [47]. Plasma homocysteine levels in adolescents are elevated in DKA and gradually normalize after insulin initiation [41]. The relatively rapid homocysteine rise may result in the very gradual normalization of protein C activity that has been observed. In adolescents, Protein S antigen levels remain normal during DKA while free protein S, the active anticoagulant, is reduced and does not return to baseline with treatment [41]. This is consistent with the finding that low levels of free protein S are the result of increased levels of C4b-binding protein in poorly controlled adult T1DM patients [30].
vWF is synthesized and secreted by endothelial cells, facilitates platelet adhesion, and is a carrier protein for factor VIII. High vWF levels are a marker of endothelial injury and activation. In adolescents, vWF antigen and activity are initially increased in DKA and decrease slowly with DKA therapy [41]. Factor VIII concentration is also elevated in adults with long-standing insulin-dependent diabetes mellitus during an episode of DKA [48,49]. The fibrinogen concentration remains normal throughout DKA and its management [41]. However, fibrinogen circulating in an environment of high glucose can become hyperglycosylated [50] with resulting fibrin fibers that are resistant to plasmin degradation [51].
In 8 adults receiving a continuous subcutaneous insulin infusion, researchers examined the effects of infusion cessation for 4 hours [40]. All subjects entered early biochemical DKA. Tests of fibrinolytic activity after vascular stimulation    Combined, these findings suggest that impaired fibrinolytic activity is an early event in DKA. Furthermore, tissue factor (TF), the primary initiator of coagulation, was significantly increased as soon as the insulin infusion was halted. No other coagulation factor demonstrated altered serum levels, and it is not yet clear that TF promotes procoagulant changes in diabetic patients. An examination of 34 adult patients with uncomplicated DKA demonstrated evidence of endothelial injury, platelet activation, relative hypofibrinolysis, and activation of the coagulation system, even in the absence of clinical signs of thrombosis [35]. Not surprisingly, multivariate analysis indicated that many of the endothelial, clinical, and hemostatic factors were interrelated. Unlike the trial of insulin infusion cessation [40], Ileri and colleagues found that fibrinolytic activity (tPA and plasmin-α 2 -antiplasmin complex levels) was increased both before and during DKA treatment. However, the upregulation was not to a degree expected for the increase in coagulation activity (thrombin-antithrombin III complex and prothrombin fragment 1 + 2 levels) at DKA presentation. Not all alterations in the coagulation system are procoagulant. Antithrombin III activity is generally increased in T1DM. In contrast, at DKA presentation, levels are slightly lower than baseline but are still higher than in a control population [35].

Platelet Numbers and Function.
Increased platelet activity has been inconsistently demonstrated in children with stroke [5], though no systematic studies of platelet activity and aggregation have been performed in children during DKA. Although platelet counts are generally normal in T1DM, platelet function is enhanced both chronically [38] and during an episode of DKA. Adult volunteers experiencing acute hyperglycemia following an oral glucose challenge demonstrate an acute increase in platelet aggregability [39]. It has been postulated that increased platelet activity may be related to decreased nitric oxide availability reported during episodes of DKA [52,53]. Ileri and colleagues found that platelet activation coexisted with DKA and was completely normalized after recovery [35].

Blood Volume, Flow, and Vascular
Reactivity. It has been demonstrated in other conditions that dehydration alone does not account for hypercoagulability [54]. Therefore, although DKA may result in significant fluid losses, other factors such as coagulation system abnormalities, hyperglycemia, acidemia-induced red blood cell rigidity (increased blood viscosity) [55,56], and vasoconstriction induced by hyperglycemia may all have an additive role. The vascular response in hyperglycemia has generally been considered vasoconstrictive, and there is some evidence that this may be related to decreased availability of nitric oxide [53]. However, vascular endothelial growth factor expression is increased by circulating ketones (β-hydroxybutyrate), leading to activity of nitric oxide-guanylate cyclase pathway, and therefore vasodilation, in mouse models [57]. Clinically, a transcranial Doppler ultrasound study of 5 children with DKA demonstrated significant vascular dysregulation with vasodilation, decreased cerebral blood flow velocity, and loss of normal cerebral blood flow regulation that only normalized after treatment [58]. Another group found normal to increased cerebral blood flow with impaired cerebral autoregulation during episodes of DKA not associated with overt CE in 6 children [59]. Importantly, none of these studies were able to define the effects that DKA has on local microvascular tone and regulation, and therefore on cerebral oxygen delivery.

Pathogenesis of Hemorrhagic Stroke
While much of the above discussion has focused on the pathogenesis of thrombosis, it is reasonable to suppose that the pathophysiology of hemorrhagic stroke may involve similar principles to that of hypoperfusion or thrombotic stroke. During an episode of DKA, hemorrhage risk is increased by endovascular perturbation secondary to the proinflammatory state, and to hyperglycemia and acidosis causing oxidative injury [24,30,31], as well as ischemic injury to the vessels from cerebrovascular dysregulation [58,59] or presentation in a shock state [24,25]. In a case-control study of 41 adult patients with hemorrhagic stroke, 31% of diabetics had hemorrhagic conversion of infarcts compared to 18% of nondiabetic stroke patients [60]. While these were not in the setting of DKA, it does raise the possibility that the aforementioned chronic inflammatory state that exists in diabetes mellitus [27,36] places these patients at higher risk of hemorrhagic conversion of ischemic brain infarction.
DIC has been reported in children with DKA [33]. The consumption of coagulation factors in DIC may predispose to hemorrhage. Four cases of isolated intracerebral DIC, including one case of a 3-month old ( Table 1, patient 1), have been reported [13,34]. DIC was identified on postmortem examination as wide-spread occlusion of small vessels by thrombus, many of which were surrounded by petechial hemorrhage. Furthermore, vascular malformations such as arteriovenous malformations (AVMs), aneurisms, and cavernous malformations that predispose to stroke may be relevant in this patient group. In children who experience hemorrhagic stroke in the absence of diabetes mellitus, vascular malformations are the most commonly experienced risk factor, occurring in 20-85% in case series [61].
DKA is a proinflammatory condition with vascular endothelial perturbation, and dysregulation of the coagulation system features associated with abnormal levels and activities of several coagulation factors, including platelets, which result in a procoagulant state. Additional factors contributing to the procoagulant state are abnormalities in blood volume, blood viscosity, cerebral autoregulation and blood flow, while vascular injury may be a result of oxidative injury and ischemia related to systemic hypoperfusion, vascular dysregulation, or cerebral edema. Hemorrhagic stroke likely arises secondary to hypoxia and the vascular injury encountered in the oxidative, proinflammatory state of DKA.

Pathology: Ischemic versus Hemorrhagic Stroke
We identified cases of stroke associated with DKA in children and youth through an exhaustive search of the literature. We searched PubMed for the following terms: ["stroke" or "brain infarction" or "thrombosis"] and ["DKA" or "diabetic ketoacidosis"]. Titles were hand searched for relevance, and where significance was unclear the abstract was read. All appropriate articles were obtained and their reference lists were scanned for other articles of relevance. This iterative process was continued until no further new reports became apparent. We have presented (Table 1) the reported cases of DKA-associated stroke in the following categories: arterial ischemic stroke [4][5][6][14][15][16], cerebral venous thrombosis [16][17][18][19][20][21], and hemorrhagic stroke [4,15,[22][23][24][25]. Based on the limited investigative modalities available to those compiling the reports, it is likely that our ability to differentiate between stroke etiologies (e.g., arterial ischemia due to hypoperfusion or thrombus, or hemorrhagic stroke with bleeding that arose de novo versus secondary to bleeding within an ischemic injury) is limited. This was clearly illustrated in the case control series described by Muir and colleagues, in which 4 of 23 children with DKA who had CT scanning after developing clinical signs of cerebral edema demonstrated subarachnoid or intraventricular hemorrhage without radiologic edema [62]. Thus we determined that, where hemorrhage was the only radiologic finding, the patient was included in the "hemorrhagic stroke" group. The 4 patients from the latter study were not included in Table 1, as there was no specific individual information given. Where multiple small hemorrhages occurred in the setting of clear thrombus or emboli [13], the patient was included in the "arterial ischemic stroke" group. Cases of venous sinus thrombosis with relevant clinical findings in the acute phase, but with no long-term clinical or radiologic sequelae [16], were included, as they represent acute intracerebral events associated with DKA.
The pathologic tissue findings of a patient having experienced acute cerebral infarction related to an episode of DKA are not expected to be different from those of a nondiabetic child who has had a stroke. Two patients (patients 26 and 27) had hemorrhagic stroke that was characterized histologically by a ring-and-ball morphology ( Figure 1) [24]. Widespread small vessel occlusion, diagnosed on autopsy, has also been described [13]. As stroke itself may cause cerebral edema, it becomes difficult to ascertain whether cerebral edema in DKA is the cause or an effect of acute cerebral infarction. Table 1 details the course of four patients (2,3,9,24) who had signs of raised intracranial pressure and may have suffered cerebral hypoperfusion and infarction as a complication of DKA-CE. Patients 2, 3, 6, 8, and 28 had infarction limited to areas supplied by the posterior cerebral artery and anterior choroidal arteries (the areas most susceptible to damage following transtentorial herniation), presumably secondary to raised intracranial pressure. However, some of this group lacked clinical signs or radiological evidence (patients 6,8,28) of CE, raising the possibility of a primary thrombotic or hemorrhagic event. Patient 1 had signs of DKA-CE but also clear thrombus formation on autopsy [13].
Two patients were found to have thrombophilic conditions. Patient 14 developed a DVT prior to presentation with central nervous system complications of DKA. MRI demonstrated acute infarctions in multiple areas of the brain, and examination identified heterozygosity for Factor V Leiden [18]. Patient 18, who developed thrombosis of multiple cerebral venous sinuses, was found to have heterozygous mutation of the prothrombin gene (G20210A) [21].

Evaluation
The clinical presentation of stroke as a focal neurologic deficit should pose little diagnostic problem. Of greater difficulty is differentiating global neurological impairment in DKA patients from severe acidosis, DKA-CE, or primary stroke. Among the patients presented in Table 1, only 8 (patients 10,11,12,15,17,23,25,28) had focal neurologic signs. The remainder presented with nonspecific signs consistent with the global dysfunction seen in DKA-CE. As the presentation of CE and primary stroke in DKA can be so similar, it is imperative that the clinician have a high index of suspicion for stroke. Early imaging is warranted once the patient is stabilized in order to optimize management.
The best modality for identification of the ischemia associated with stroke is magnetic resonance imaging (MRI) with perfusion-and/or diffusion-weighted imaging, which have sensitivity nearing 100% [63]. Computed tomography (CT) may be used to rule out CE, hemorrhage or abscess, though the sensitivity for identification of ischemic infarction in the acute phase is only 50% [63]. Additionally, it is agreed that CT may miss cases of cerebral edema, though its sensitivity for identifying elevated intracranial pressure has been reported at 99.1% [64]. Cerebral angiography is the gold standard for assessment of the cerebral vasculature, although MR angiography (MRA) is able to detect large vascular lesions effectively [65], with the benefit of being noninvasive. CT angiography, though it requires injection of contrast, may be used to evaluate the cerebral circulation early in stroke evolution [66]. In cases where neither MRA nor MRI defines suspected pathology in a distal artery, cerebral angiography may be considered.
All patients with suspected intracranial pathology should have a coagulation screen performed. In cases of clear Stroke Research and Treatment 9 thrombosis or hemorrhage, more detailed analysis of the hemostatic system is warranted. Identification of the exact histopathology may be done with tissue biopsy [24], though this carries many risks and is not recommended as a matter of course.

Management
Fully evidence-based management guidelines for children experiencing acute ischemic or hemorrhagic infarction do not exist and have been extrapolated from adult data. Admission to the critical care unit and close monitoring is appropriate for any patient with suspected or proven central nervous system (CNS) complication of acute DKA. Unless diffuse CE can be absolutely excluded, or another clear cause is present, emergency management for CNS complications of DKA should prioritize the treatment for cerebral edema. Although early reactive treatment for CE appears beneficial [4,67], it was stated in 1990 that intracerebral complications do not often come with warning signs, and early intervention measures are frequently unsuccessful in preventing complications; so prevention of DKA is the most effective method of preventing complications [4]. In the intervening 20 years, this generally remains true, although there may be some alternative management options for thrombus causing stroke.
Children with stroke should receive aggressive treatment for fever, infection, and seizures [68]. For all forms of stroke, recommendations are for early mobilization and rehabilitation. Current guidelines do not support the use of thrombolytics in pediatric arterial ischemic stroke [69], although thrombotic stroke associated with an episode of DKA is not addressed specifically. Furthermore, multiple case reports document use of thrombolytics treatment associated with good outcome in children with acute ischemic cerebral infarction [70][71][72], including thrombolysis used successfully up to 36 hours after onset of symptoms [73,74]. Safe, acute thrombolysis with recombinant tPA was reported for a 10year-old child with cerebral venous sinus thrombosis that occurred 3 days after onset of DKA, who had complete recovery [21]. However, caution is warranted as the risks of thrombolysis and the optimal tPA doses in children have not been quantified. In light of the lack of evidence or strength of recommendations, it seems prudent that management of arterial ischemic stroke in association with DKA be considered on a case-by-case basis and in consultation with stroke experts.
Beyond the acute phase of ischemic or thrombotic stroke it has generally been agreed that anticoagulation with heparin [75,76] may be appropriate for pediatric patients who have already experienced arterial ischemic stroke or cerebral venous thrombosis, independent of T1DM and DKA. International guidelines recommend the use of low molecular weight or unfractionated heparin initially, followed by warfarin therapy for 3-6 months [75][76][77]. There is no clear guidance on the management of children who have experienced hemorrhagic stroke. Adult guidelines suggest that extremely high blood pressures be reduced cautiously and recombinant factor VIIa has shown some promise in decreasing recurrence but is only recommended within clinical trials. However, these guidelines refer to different etiologies for intracranial hemorrhage than are seen in children or in the DKA population. Large hemorrhages compromising neurovascular structures may warrant surgical decompression.
Prophylactic systemic anticoagulation has been suggested for patients in DKA [13,35,78]. However, anticoagulation is not addressed in international consensus statements on the management of DKA [1,79,80]. In light of the risk of hemorrhagic stroke and the unknown incidence of stroke in pediatric DKA, a broad recommendation for prophylactic anticoagulation cannot be supported at this time.

Outcome
Stroke outcome depends on the cerebral regions affect and the extent of the injury. As demonstrated from the case series (Table 1), the majority of children with DKA-associated stroke reported have some form of residual neurologic deficit with death or persistent vegetative state as the outcome in 8 of 28 patients (29%) and full recovery seen in only 4 of 28 cases (14%).

Conclusions
Stroke in DKA is uncommon but life-threatening. DKA may be considered an inflammatory condition with vascular endothelial perturbation and dysfunction of the coagulation system. Multiple causes of thrombus have been postulated and studies show several contributing mechanisms. Hemorrhagic infarctions are rare and may be multifactorial but must be considered a risk. Management for CNS complications of DKA should prioritize the treatment for cerebral edema. As the initial presentation of pediatric stroke can be subtle and may be confused with DKA-CE, early imaging for any young person with neurologic deterioration in association with an episode of DKA is imperative following emergency treatment for CE and stabilization.