Seizures after intracerebral and subarachnoid hemorrhage. An 82-year-old woman presented with sudden-onset headache, dysphasia, and right hemiparesis while receiving anticoagulation therapy for chronic atrial fibrillation. Computed tomographic (CT) scan at admission (A) demonstrates an acute left temporal lobe intraparenchymal hematoma, with adjacent subdural and subarachnoid hemorrhage. The responsible lesion, an aneurysm of the middle cerebral artery, was treated surgically. During the postoperative period, she had episodes of right-sided facial twitching associated with transient worsening of her aphasia. A postoperative CT scan 3 months later (B) showed a hypodense lesion of the midtemporal lobe in the middle cranial fossa. Electroencephalographic (EEG) findings (C) demonstrated focal spike waves, consistent with seizure activity, followed by focal slowing and periodic lateralizing epileptiform discharge in the left hemisphere. Ref indicates reference.
Mass lesion causing focal seizures. A 43-year-old woman presented to the emergency department after focal-onset right-sided clonic movements, followed by loss of consciousness. Computed tomographic scan with contrast demonstrated a giant aneurysm of the middle cerebral artery with adjacent cerebral edema, contrast uptake into the aneurysm, and an adjacent region in which a thrombus has formed. The patient underwent successful craniotomy, with thrombectomy and aneurysm clipping.
Status epilepticus due to the reperfusion syndrome. A 66-year-old woman underwent an otherwise uncomplicated right carotid endarterectomy for asymptomatic extracranial carotid artery disease 3 days earlier. She awoke with a headache, followed by left-arm clonic movements. These progressed to generalized, tonic-clonic seizures, which were not aborted by administration of lorazepam and phenytoin sodium therapy. The continuous electroencephalographic (EEG) monitor showed periodic lateralized epileptiform discharges occurring every 2 to 5 minutes, lasting a few seconds (A). Subsequent T2- and diffusion-weighted magnetic resonance imaging studies of the brain showed right-sided thalamic (B) and cortical lesions (C). Phenobarbital sodium was then administered, which arrested seizure activity. Ref indicates reference.
Silverman IE, Restrepo L, Mathews GC. Poststroke Seizures. Arch Neurol. 2002;59(2):195-201. doi:10.1001/archneur.59.2.195
Stroke is the most common cause of seizures in the elderly, and seizures are among the most common neurologic sequelae of stroke. About 10% of all stroke patients experience seizures, from stroke onset until several years later. This review discusses current understanding of the epidemiology, pathogenesis, classification, clinical manifestations, diagnostic studies, differential diagnosis, and management issues of seizures associated with various cerebrovascular lesions, with a focus on anticonvulsant use in the elderly.
In population studies, stroke is the most commonly identified cause of epilepsy in adult populations older than 35 years.1 In the elderly, stroke accounts for more than half of the newly diagnosed cases of epilepsy in which a cause is determined, ahead of degenerative disorders, brain tumors, and head trauma.1 From stroke registry data, about 5% to 20% of all individuals who have a stroke will have subsequent seizures,2,3 but epilepsy (recurrent seizures) will develop in only a small subset of this group. Given that, in each year, more than 730 000 people in this country have a stroke, a conservative incidence of seizures after stroke is about 36 500 new cases per year.
The largest and most rigorous methodological attempt to examine poststroke seizures was the prospective multicenter report from the Seizures After Stroke Study Group.2 The study enrolled 1897 patients and found an overall incidence of seizures of 8.9%. Recurrent seizures consistent with the development of epilepsy were rare, occurring in 2.5% of the patients, but the mean follow-up was only 9 months.
Seizures may be a more common accompaniment of hemorrhagic rather than ischemic stroke.
Bladin et al2 found the incidence of seizures to be 10.6% among 265 patients with intracerebral hemorrhage vs 8.6% among 1632 with ischemic stroke. In another prospective series,4 seizures occurred in 4.4% of 1000 patients, including 15.4% with lobar or extensive intracerebral hemorrhage, 8.5% with subarachnoid hemorrhage, 6.5% with cortical infarction, and 3.7% with hemispheric transient ischemic attacks. A seizure was the presenting feature of intracranial hemorrhage in 30% of 1402 patients.5 Among 95 patients with aneurysmal subarachnoid hemorrhage, the incidence rate of prehospital seizures was higher (17.9%) than that occurring in the hospital (4.1%).6
Seizures after stroke are classified as early or late onset, according to their timing after brain ischemia, in a paradigm comparable to post-traumatic epilepsy.2,7 An arbitrary cut point of 2 weeks after the presenting stroke has been recognized to distinguish between early- and late-onset poststroke seizures.5,8,9 Different characteristics and mechanisms of poststroke seizures, according to their proximity to the onset of brain ischemia, have been proposed, but no clear pathophysiological basis exists for the 2-week cut point.
Most early-onset seizures occur during the first 1 to 2 days after ischemia. Almost half (43%) of all patients in the Stroke After Seizures Study experienced a seizure within the first 24 hours after stroke.2 In a series restricted to early-onset seizures, 90% of the 30 patients had ictal activity within the first 24 hours.10 Most seizures associated with hemorrhagic stroke also occur at onset or within the first 24 hours.11
During acute ischemic injury, accumulation of intracellular calcium and sodium may result in depolarization of the transmembrane potential and other calcium-mediated effects. These local ionic shifts may lower the seizure threshold.2,12 Glutamate excitotoxicity is a well-established mechanism of cell death in the experimental stroke model. Antiglutamatergic drugs may also have a neuroprotective role in ischemic settings, aside from the role of treating seizures.
The size of regional metabolic dysfunction may also be relevant in causing early-onset seizures. In the setting of large regions of ischemic hypoxia, high levels of excitotoxic neurotransmitters may be released extracellularly. In studies of the postischemic brain in experimental animal models, neuronal populations in the neocortex13 and hippocampus14 have altered membrane properties and increased excitability, which presumably lower the threshold for seizure initiation. The ischemic penumbra, a region of viable tissue adjacent to the infarcted core in ischemic stroke, contains electrically irritable tissue that may be a focus for seizure activity.
In addition to focal ischemia, global hypoperfusion can cause seizure activity. Hypoxic-ischemic encephalopathy is one of the most common causes of status epilepticus and carries a poor prognosis. Particularly vulnerable to ischemic insult is the hippocampus, which is an especially epileptogenic area.
In late-onset seizures, by contrast, persistent changes in neuronal excitability occur. Replacement of healthy cell parenchyma by neuroglia and immune cells may play a role in maintaining these changes. A gliotic scarring has been implicated as the nidus for late-onset seizures, just as the meningocerebral cicatrix may be responsible for late-onset post-traumatic epilepsy.2
An underlying permanent lesion appears to explain the higher frequency of epilepsy in patients with late than early-onset seizures. As in post-traumatic epilepsy,15 late occurrence of a first seizure appears to carry a higher risk for epilepsy. In patients with ischemic stroke, epilepsy developed in 35% of patients with early-onset seizures and in 90% of patients with late-onset seizures.16 The risk for epilepsy was comparable in patients with hemorrhagic stroke; epilepsy developed in 29% of patients with early-onset seizures vs 93% with late-onset seizures.5
The concept that cardiogenic emboli to the brain are more likely to cause seizures acutely is controversial, with few supporting data. Among 1640 patients with cerebral ischemia,17 events attributed to a cardiac source were most commonly associated with early-onset seizures (16.6%), even compared with supratentorial hematomas (16.2%). However, the definition of cardiogenic mechanism in this series was often based on nonspecific criteria. Several authors have questioned the association of seizures with cardioembolic events.3,4 Seizures at onset were not a criterion in a data bank study of the cardiac causes of stroke.18 Intuitively, there is no reason to suspect that cardioembolic lesions would be more likely than emboli from large-vessel sources to cause seizures, as cardiac and large-vessel emboli frequently involve lesions to distal cortical branches. The mechanism by which cortical emboli precipitate seizures is uncertain,12 but possibilities include depolarization within an ischemic penumbra, rapid reperfusion after the fragmentation and distal migration of the embolus,19 or a combination of both.
Cortical location is among the most reliable risk factors for poststroke seizures.2 Poststroke seizures were more likely to develop in patients with larger lesions involving multiple lobes of the brain than in those with single lobar involvement.20 However, any stroke, including those with only subcortical involvement, may occasionally be associated with seizures.20 Earlier studies, relying on less sensitive neuroimaging techniques, may not have detected concomitant small cortical lesions that could cause ictal activity. The mechanism by which deep hemispheric subcortical lesions, most commonly due to small-vessel disease, cause seizures is not understood.2
Analogous to cortical involvement in ischemic stroke, a lobar site is considered to be the most epileptogenic location in patients with intracerebral hemorrhage. In a series of 123 patients,21 seizure incidence was highest with bleeding into lobar cortical structures (54%), low with basal ganglionic hemorrhage (19%), and absent with thalamic hemorrhage. Caudate involvement of the basal ganglia and temporal or parietal involvement within the cortex predicted seizures.21 Hemorrhage due to cerebral venous thrombosis also commonly presents with seizures. Parenchymal, often cortical, hemorrhage resulting from local venous congestion is the likely cause of seizure activity.
The mechanism of seizure initiation by hemorrhage is not established. Products of blood metabolism, such as hemosiderin, may cause a focal cerebral irritation leading to seizures, analogous to the animal model of focal epilepsy produced by iron deposition on the cerebral cortex.22 In subarachnoid hemorrhage, there is often extensive hemorrhage into the basal cisterns, which directly contacts the frontal and temporal lobes. Patients with subarachnoid hemorrhage also may have an intraparenchymal component to the hemorrhage (Figure 1).
The only clinical predictor for seizures after ischemic stroke is the severity of the initial neurologic deficit. Greater initial stroke severity9 or stroke disability2 predicted seizures. By contrast, in the Oxfordshire Community Stroke Project, only 3% of 225 patients who were independent 1 month after a stroke experienced a seizure between 1 month and 5 years.23 Patients presenting with greater neurologic impairment tend to have larger strokes that involve wider cortical areas.
In retrospective studies, risk factors for seizures after subarachnoid hemorrhage included middle cerebral artery aneurysms,24 intraparenchymal hematoma,25 cerebral infarction,26 a history of hypertension,27 and thickness of the cisternal clot.6 By contrast, clinical predictors for seizures after intraparenchymal hemorrhage have been lacking.2
Vascular lesions may cause seizures by other mechanisms. Seizures due to arteriovenous malformations and aneurysms typically occur when these lesions rupture, but these vascular lesions may cause seizures by directly irritating adjacent brain parenchyma (Figure 2).
Finally, seizures associated with vascular lesions occur in the setting of significant reperfusion after revascularization procedures, most commonly carotid endarterectomy for chronic severe extracranial carotid stenosis. The reperfusion syndrome, first described by Sundt and colleagues,28 includes transient focal seizure activity, atypical migrainous phenomena, and intracerebral hemorrhage, although the clinical triad is often incomplete. Onset of this rare syndrome ranges from several days to 3 weeks after revascularization29 and often is signaled by a new ipsilateral headache.30 Surgical correction of an arteriovenous malformation may also cause intraoperative or postoperative hyperemia, with subsequent seizures or hemorrhage.19,31 By contrast, arteriovenous malformations located in border-zone regions subject to relatively low flow rates have a lower risk for hemorrhage.32
The reperfusion syndrome has been attributed to impaired cerebral autoregulation.19 In the setting of chronic hypoperfusion due to high-grade carotid stenosis, the arterioles responsible for normal autoregulation in the downstream cerebral hemisphere become chronically dilated. Subsequently, when perfusion is improved by a revascularization procedure, the vessels are incapable of vasoconstriction, and the brain parenchyma is subjected to a massive augmentation of blood flow. The release of vasoactive neuropeptides from perivascular sensory nerves may contribute to the development of the reperfusion syndrome,19 to oxidants that develop before revascularization,33 and to an inflammatory response to the reestablishment of circulation.34
Given that most poststroke seizures are caused by a focal lesion, poststroke seizures are typically focal at onset. In a study of early-onset seizures in 90 patients,17 simple partial seizures were the most common type (61%), followed by secondarily generalized seizures (28%). In other series,3,10 early-onset seizures were more likely to be partial, whereas late-onset seizures were more likely to generalize secondarily. Most recurrent seizures are of the same type as the presenting episode, and they tend to recur within 1 year on average.
In a large series of patients with poststroke seizures, 9% had status epilepticus.35 The only associated finding was higher functional disability; status epilepticus was not associated with a higher mortality rate, stroke type (ischemic or hemorrhagic), topography (cortical involvement), lesion size, or electroencephalographic (EEG) patterns.
The phenomenological features of the reperfusion syndrome are similar in that focal onset with occasional secondary generalization is the rule. Seizure activity always occurs in the ipsilateral vascular territory of the surgically treated internal carotid artery.28 Occasionally, status epilepticus ensues (Figure 3).
Holmes36 found that patients with periodic lateralizing epileptiform discharges and bilateral independent periodic lateralizing epileptiform discharges on EEG after stroke were especially prone to the development of seizures. Those patients with focal spikes also had a high risk of 78%. Focal slowing, diffuse slowing, and normal findings on EEG, by contrast, were associated with relatively low risks of 20%, 10%, and 5%, respectively. Other work found that cortical involvement on results of neuroanatomical imaging studies was more predictive of epilepsy than any single EEG finding.10
Focal slowing on EEG may simply reflect a wide region of ischemic or infarcted tissue involving the cerebral cortex or subcortical territory. In addition to neuroimaging, EEG may be helpful in the early evaluation of poorly defined poststroke focal neurologic symptoms. In selected patients, focal slowing may confirm the clinical impression of hemispheric ischemia and argue against ongoing seizures as an explanation for an acute neurologic syndrome. The absence of EEG abnormalities does not definitively exclude cerebral ischemia, particularly in subcortical or subtentorial structures, or intermittent seizure activity.
Uncommonly, seizures may mimic ischemia and infarction on neuroimaging findings. Lansberg et al37 recently described several acute magnetic resonance imaging findings in 3 patients with partial status epilepticus. These studies showed cortical hyperintensity on diffusion-weighted imaging and T2-weighted sequences and a corresponding area of low apparent diffusion coefficient. However, these findings were readily differentiated from typical ischemic stroke in their nonvascular distributions, increased signal of the ipsilateral middle cerebral artery on magnetic resonance angiography, and leptomeningeal enhancement on postcontrast magnetic resonance imaging. Another study showed a hyperintensity on diffusion-weighted images in the dorsolateral portion of the ipsilateral thalamus in 2 patients (Figure 3).38
The differential diagnosis of ischemia-induced seizures includes secondary seizures due to other causes. Medications, drug therapy withdrawal (eg, benzodiazepines), and metabolic disturbances (eg, glucose abnormalities) typically cause generalized seizures, unless an underlying lesion is already present. Migraine-related focal phenomena and transient ischemic attacks may produce focal slowing on EEG findings. Among these entities, glucose abnormalities should not be overlooked.
Choice of an anticonvulsant medication should be guided by the individual characteristics of each patient, including concurrent medications and medical comorbidities. To our knowledge, no controlled trials evaluating only poststroke seizures have been conducted to assess specific agents. Perhaps a more important issue is whether to initiate treatment at all, because only a few poststroke seizures have been demonstrated to recur. In the absence of absolute predictors of poststroke epilepsy, most physicians empirically treat patients for their seizures when they occur in the setting of a recent stroke. Thus, Bladin et al2 argue that a controlled trial including patients with poststroke epilepsy would pose extensive logistical challenges and would likely be unethical. Despite the relatively low incidence of poststroke seizures, the absolute numbers are still high, arguing that an important problem exists. Arboix et al39 concluded that the efficacy of anticonvulsant prophylaxis should be assessed in prospective, randomized trials conducted with high-risk patients.
Poststroke seizures are usually well controlled with a single anticonvulsant. In 1 retrospective study, seizures in 88% of the 90 patients could be managed with monotherapy.10 Given the typical focal onset of poststroke seizures, first-line therapy options include carbamazepine and phenytoin sodium. The latter has the advantage of parenteral administration, which may be necessary given that swallowing or mental status likely will be impaired. Fosphenytoin sodium is also a noteworthy option in stroke patients because of lesser cardiotoxicity than phenytoin. Benzodiazepines, particularly lorazepam, should be initially administered to the patient with ongoing seizures. No data support the use of different agents to treat early vs late-onset seizures.
The newer antiepileptic drugs are being touted as first-line agents for elderly patients because of their effectiveness and favorable side-effect profiles. Approximately 10% of nursing home residents in the United States receive antiepileptic drugs, most often for the treatment of seizure disorders.40 In a trial of newly diagnosed epilepsy in the elderly, lamotrigine was recently demonstrated to be better tolerated and to maintain patients free of seizures for longer intervals than carbamazepine.41 Although many of the newer anticonvulsants, eg, topiramate42 and levetiracetam,43 have been studied as adjunctive agents for refractory partial seizures, in practice they are often used as monotherapy. Gabapentin has been shown to be efficacious as monotherapy for partial epilepsy.44 For all antiepileptic drugs, the chief relevant dose-limiting adverse effect is sedation, particularly in the elderly stroke patient.
Drug interactions are an important consideration, since most stroke patients take multiple medications.40 The first-generation antiepileptic agents undergo significant hepatic metabolism, and phenytoin and valproic acid are highly protein bound. For example, the well-recognized interaction of warfarin with phenytoin makes it difficult to maintain consistent therapeutic ranges of both agents.
In its guidelines, the Stroke Council of the American Heart Association45,46 recommended uniform seizure prophylaxis in the acute period after intracerebral and subarachnoid hemorrhage. For intracerebral hemorrhage, seizure activity may result in further neuronal injury and contribute to coma, although no clinical data support this recommendation.45 Patients with solely cerebellar or deep subcortical (eg, thalamic) lesions are at very low risk for seizures and need not be treated. The guideline suggests a dose of phenytoin sodium titrated by serologic levels (14-23 µg/mL), with discontinuation of therapy after 1 month if no seizure activity occurs during treatment.47 Patients with seizure activity more than 2 weeks after presentation are at a higher risk for recurrence and may require long-term seizure prophylaxis.48
Small retrospective studies suggest no benefit of prophylactic anticonvulsants after aneurysmal subarachnoid hemorrhage.49 However, because of the relatively low risk associated with antiepileptic therapy and the great concern about aneurysmal rebleeding, a clinical trial addressing this issue may never occur. Long-term use of antiepileptic agents is not recommended for patients with subarachnoid hemorrhage who do not have seizures, but should be considered when at least 1 of several risk factors is present.46
In the case of reperfusion syndrome, the critical preventive measure is likely the aggressive control of systemic blood pressure.19 The role of antiepileptic therapy in this population of patients is unclear. Seizures during the reperfusion syndrome sometimes respond to antiepileptic drugs, according to anecdotal evidence,50 but may be difficult to treat in the absence of heavy sedation. Some surgeons empirically administer prophylaxis owing to the concern about seizures for 1 to 2 weeks after endarterectomy in patients with high-grade carotid stenosis. Management of venous infarction often dictates administration of systemic anticoagulation, and, recently, intrathrombus thrombolysis via endovascular delivery has demonstrated success in selected patients.51 Authorities recommend instituting antiepileptic therapy only if seizures occur.52
The impact of poststroke seizures on stroke outcomes is unclear, with conflicting data from different case series. In 2 prospective studies, early-onset seizures were not associated with higher mortality rates9 or worse neurologic deficits.4 Seizures were associated with better outcomes on the Scandinavian Stroke Scale in another series9; the authors postulated that seizures were a manifestation of a larger ischemic penumbra that contributed to better recovery. Conversely, in other work, patients who had early-onset seizures within 48 hours of the presenting stroke or transient ischemic attack were significantly more likely to die in the hospital (37.9%) vs those who did not present with seizures (14.4%).39
For subarachnoid hemorrhage, seizures at onset predicted late-onset seizures and poor outcomes, measured by disability 6 weeks later using the Glasgow Outcome Scale.53 In an Icelandic population, epilepsy was more frequent in patients with severe neurologic residua (48%) compared with those without (20%).8
Seizures in the reperfusion syndrome are usually self-limited. Long-term prognosis depends on the development of intracerebral hemorrhage. In a series of 1500 carotid endarterectomies, hemorrhage developed in 11 patients, and 4 of these patients died.54
Poststroke seizures are a common and treatable phenomenon, whereas the development of epilepsy is relatively rare. Cerebrovascular lesions associated with the development of seizures include the following: intracerebral (parenchymal) and subarachnoid hemorrhage and cerebral venous thrombosis, with or without venous infarction; lesions involving the cerbral cortex; larger neurologic deficits or disability at presentation; and revascularization procedures involving the extracranial internal carotid artery. The treatment of poststroke seizures is no different than the approach to treatment of partial-onset seizures due to other cerebral lesions, and poststroke seizures usually respond well to a single antiepileptic drug. Given their tolerability, the newer generations of anticonvulsant agents hold promise in treating older patients. Given the low incidence of poststroke epilepsy, there is no indication for seizure prophylaxis in patients with acute ischemic stroke who have not had a well-documented first event. The need for chronic anticonvulsant use should be evaluated periodically, perhaps every 6 months. Despite the absence of clinical data documenting effectiveness, most patients presenting with intracerebral or subarachnoid hemorrhage should receive short-term antiepileptic prophylaxis.45,46
Future areas of research regarding poststroke seizures include assessing their impact on initial lesion size and on delayed patient outcomes, determining the appropriateness of chronic antiepileptic therapy after a single seizure, and establishing risk factors for the reperfusion syndrome. Poststroke epilepsy may also become an important basic model in research that aims to prevent the transformation of injured cerebral tissue into an epileptic focus.
Accepted for publication September 25, 2001.
Author contributions: Study concept and design (Drs Silverman, Restrepo, and Mathews); acquisition of data (Dr Silverman); analysis and interpretation of data (Dr Silverman); drafting of the manuscript (Drs Silverman and Restrepo); critical revision of the manuscript for important intellectual content (Drs Silverman, Restrepo, and Mathews); administrative, technical, and material support (Drs Silverman and Restrepo); study supervision (Dr Silverman).
We thank Richard H. Mattson, MD, for his critical review of the manuscript.
Corresponding author and reprints: Isaac E. Silverman, MD, The Stroke Center at Hartford Hospital, 85 Seymour St, Suite 800, Hartford, CT 06106 (e-mail: email@example.com).