Epilepsy after brain insult

Measurement of epilepsy risk factors.

The importance of a particular risk factor for the development of unprovoked seizures can be assessed in several ways. Cohort studies identify patients with a brain insult of interest and follow them over time for the development of seizures. The relative risk measures the magnitude of the association between the risk factor and seizures by comparing the seizure incidence rates of subjects with the risk factor to those without it (usually the general population). Risk factors with relative risks greater than 10 have a strong causal association with seizures, and those between 4 and 10 have a probable causal association (figure 1). The cumulative incidence (table) measures the probability that an individual will develop seizures within a defined time period after various types of brain injury. Risk factor impact can also be assessed by the attributable risk, or the percentage of seizure or epilepsy cases that can be ascribed to the risk factor of interest (figure 2). Attributable risk can be considered as the percentage of epilepsy cases that could be prevented if the risk factor were completely eliminated. In adults, a probable etiology can be determined for almost 50% of new-onset seizures. The most common risk factors are cerebrovascular diseases (21%), tumors (11%), and traumatic brain injury (7%).2 Such brain insults with high attributable risk for epilepsy are attractive targets for antiepileptogenesis, because prevention of seizures after such injuries would result in a significant decrease in the overall incidence of epilepsy.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Relative risk for unprovoked seizures after common brain injuries. The dotted vertical line represents the general population risk for unprovoked seizures. MR/CP = mental retardation/cerebral palsy; SAH = subarachnoid hemorrhage; CVA = cerebrovascular accident; TBI = traumatic brain injury.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Causes of epilepsy (children and adults). The proportions of epilepsy from each type of brain injury approximate the attributable risk for that brain injury.

Seizures following brain insults.

Seizures after damage to the brain may occur acutely in the setting of a brain insult (early or acute symptomatic seizures), as a single seizure far removed from the acute insult (late, unprovoked, or remote symptomatic seizures), or as recurrent late seizures (remote symptomatic epilepsy). Acute symptomatic seizures usually occur in the first 1–2 weeks after an acute insult. The pathophysiologic mechanisms are incompletely understood, but may include changes in the blood–brain barrier, presence of parenchymal hemorrhage, release of excitotoxins such as glutamate, free radical damage, and alteration of energy metabolism.4 Acute symptomatic seizures are frequently viewed as epiphenomena of the underlying brain disorder or markers for the severity of injury, with little independent impact on outcome from injury.5-7⇓⇓ In addition, early seizures can be suppressed in several conditions, such as traumatic brain injury and intracranial surgery, by short-term prophylactic administration of antiepileptic drugs (AEDs), without altering the incidence of late epilepsy.8 Early seizures cannot be completely excluded, however, as a factor in human epileptogenesis. In animal models, acute seizures produced by a variety of mechanisms (electrical stimulation, chemoconvulsants, status epilepticus) cause progressive changes in neural networks leading to spontaneous recurrent seizures,3 and it is plausible that seizures in humans produce similar changes. Early seizures are the strongest predictor of late epilepsy following a variety of brain insults,1,9-14⇓⇓⇓⇓⇓⇓ although this effect is not clearly independent of the severity of injury. Although prophylactic treatment with AEDs decreases early clinical seizures, electrographic (subclinical) seizures continue in a high proportion of patients with acute brain insults.15 Therefore, the effect of early seizures on epileptogenesis cannot yet be accurately assessed.

Late or remote symptomatic seizures follow a latent interval of variable duration, usually weeks to months but sometimes years.11 Such seizures are believed to be the result of epileptogenesis, or chronic changes in neural networks favoring excitation. In this review, single and recurrent late unprovoked seizures are considered together. Many prospective studies do not specifically address single and recurrent late seizures, using the first late seizure as an end point, and separating these two groups is therefore often difficult. More importantly, long-term follow-up studies have revealed that the risk for recurrent seizures after a single late seizure in patients with remote brain injuries is high, usually greater than 80%.16 Because of this increased risk, many patients with a single late seizure after brain insult are started on AEDs, potentially suppressing further late seizures and masking the development of epilepsy. The “epileptogenicity” of a brain insult is therefore best approximated by evaluating the incidence of any late unprovoked seizure.

Finally, seizures may also occur as the presenting symptom or during the chronic course of a progressive neurologic disorder, such as a brain tumor or Alzheimer’s disease. Such seizures do not clearly fit with either acute or remote symptomatic seizures and may involve pathophysiologic mechanisms of both. Trials of seizure prophylaxis in these populations may therefore be difficult to generalize to the larger population of symptomatic epilepsies.

Optimal groups for trials of seizure prevention after brain insult.

Potentially modifiable risk factors for epilepsy must be identified before the onset of clinical seizures because epileptogenesis, once established, will probably be more difficult to reverse. Preventative treatment is not possible for “occult” brain lesions, such as subclinical stroke, cortical dysplastic syndromes, brain tumors, and genetic mutations, which may be silent until the occurrence of the first seizure. Populations with multifactorial epilepsy etiologies, such as cerebral palsy, mental retardation, or neurosurgical procedures, may not respond uniformly to prophylactic treatments. Optimally, the brain insult of interest should be common and should have a high association with epilepsy, to maximize the population impact of a preventative treatment. The following discussion therefore focuses on risk factors that may be good targets for neuroprotective therapy: traumatic brain injury, CNS infections, stroke, brain tumors, cerebrovascular disease, neurosurgery, Alzheimer’s disease, status epilepticus, and febrile seizures. The risk for development of late seizures, the time course of this risk, and factors influencing the likelihood of epilepsy are addressed for each of these conditions.

Traumatic brain injury (TBI).

TBI accounts for about 4% of epilepsy, and seizures after TBI are often refractory to therapy. Early seizures occur in about 15% of patients with severe TBI and in smaller proportions of those with milder injuries.17 The risk for development of late seizures is directly related to the severity of the brain injury.12 Patients with mild TBI (loss of consciousness or amnesia lasting less than 30 minutes) have a risk similar to the general population (relative risk 1.5). Moderate TBI (loss of consciousness for 30 minutes to 24 hours or a skull fracture) confers a relative risk of 2.9, but risk is elevated for only about 2 years. Survivors of severe TBI (loss of consciousness or amnesia for more than 24 hours, subdural hematoma, or brain contusion) have a relative risk of 17 compared to the general population. Almost 70% of those who ultimately develop late seizures do so within the first 2 years, but risk remains elevated for more than 5 years. More than 80% of patients with a single late seizure develop epilepsy (recurrent unprovoked seizures).16 Children with TBI have a much lower risk for late epilepsy overall, with approximately 10% incidence in severe head injury.10

Hemorrhage and cortical damage increase the risk for late epilepsy. In penetrating head injury, up to 50% of patients develop late seizures, and the risk remains elevated for life. Significant risk factors for late seizures after nonpenetrating TBI include brain contusion or parenchymal hemorrhage, subdural hematoma, skull fracture, and loss of consciousness or amnesia for more than 1 day.12 Such patients have a 2-year cumulative incidence of late seizures greater than 30%.8 Others have found that age of 65 years or older, depressed skull fracture, severity of neurologic deficit (low Glasgow Coma Score), and early seizures predict the development of late epilepsy.10,12,18,19⇓⇓⇓ Early seizures are a strong predictor of late seizures. Approximately 36% of adults with severe head injury and early seizures develop late seizures, whereas only 10% of those with severe injury but no early seizures suffer late seizures.10,18⇓ Epileptogenesis can be triggered by selective damage to vulnerable brain regions such as the hippocampus,20 by the irritating effects of intracerebral hemorrhage,21 or by cortical damage.22

TBI is a common brain insult with a high incidence of post-traumatic seizures. Epilepsy after TBI is the best-characterized symptomatic epilepsy in terms of degree of risk, subgroup analysis of those at highest risk, time course of development of seizures, and potential mechanisms of injury. Therefore, patients with moderate to severe TBI are excellent candidates for clinical trials with putative antiepileptogenic agents, and these studies should serve as models for trials in other symptomatic epilepsies.

CNS infections.

Although CNS infections clearly increase the risk for acute and late unprovoked seizures, there are few studies that quantify the degree of risk. Annegers et al.9 found that survivors of CNS infections have a three-fold increased risk for epilepsy, with a cumulative incidence of 5–10% depending on the type of infection. This risk was highest in the first 5 years after infection but remained elevated for the next 15 years. Patients with bacterial meningitis have a five-fold increased risk, mostly in the first 2 years after infection. Early or acute symptomatic seizures were a risk factor for the development of late seizures; over 20 years, 13% of those with early seizures developed epilepsy as opposed to 2% without early seizures.9 Similarly, viral encephalitis increased the risk by 16 times the general population rates for epilepsy, and the risk remained elevated for at least 15 years after infection. Those with early seizures had a 10% incidence by 5 years and a 22% incidence by 20 years, compared to 2% and 10%, respectively, in those without early seizures. Approximately 30% of patients with a bacterial cerebral abscess develop epilepsy, most within the first 5 years.23 Neurocysticercosis is the most common etiology for seizures in some developing countries,24 accounting for almost 50% of all adult-onset seizures. Many of these seizures are acute symptomatic, but calcified neurocysticercosis lesions are associated with an increased risk for development of unprovoked seizures. Although the risk for epilepsy after CNS infections is high, the fairly low population incidence of these infections would be a significant barrier to recruitment of study participants in any attempt at a prophylaxis trial.

Cerebrovascular disease.

Post-stroke seizures account for 11% of all epilepsy and 55% of newly diagnosed seizures in those older than 65 years.9 Seizures occur after 5–15% of strokes.25 Acute symptomatic seizures occur in 3–8% of patients, more commonly in those with severe cortical strokes and intraparenchymal hemorrhages.7 Late seizures occur in about 8% of patients by 10 years, but most seizures occur in the 2 years after an acute stroke.25 The risk for late seizures and epilepsy is correlated with the volume of tissue affected,26 increased stroke severity,27 and cortical location.6 Risk is not elevated with brainstem infarcts and is minimally increased with lacunar infarcts. Patients with large (total MCA or ICA) infarcts have a very high risk, at least 15% at 5 years. Although embolic strokes have been reported to increase late epilepsy, this effect disappears when stroke severity is controlled.6 Because most stroke studies do not look at both early seizures and late seizures, but only at the first seizure, the relationship between early and late seizures is difficult to determine. In one study, late seizures occurred in 32% of those with early seizures and only 10% of those without.28 In addition, late seizures occurred earlier in those who had had early seizures than those without.

Hemorrhagic strokes are associated with increased risk for late seizures. Relative risk in intracranial hemorrhage (ICH) is 11–13%, and in subarachnoid hemorrhage (SAH) is 18%.29,30⇓ Some but not all of the increased risk seen with hemorrhagic strokes disappears when stroke severity is controlled.6 One population-based study of epilepsy after SAH found that 25% of survivors developed epilepsy, all within 4 years.14 Seventy percent of those with early seizures after SAH developed epilepsy compared to only 12% of those without.

Seizures after stroke are of increasing importance as the population ages and are a growing contribution to symptomatic epilepsies. Post-stroke epilepsy incidence in some subgroups (cortical infarcts, lobar hemorrhages, SAH) approaches that of traumatic brain injury. Although less well characterized, such subgroups are good targets for clinical trials of seizure prevention.

Brain tumors.

Seizures secondary to brain tumors account for almost 4% of all epilepsy. Approximately 30–50% of patients with brain tumors present with seizures as the initial symptom, and subsequent seizures are frequently associated with a neurosurgical procedure or progression of the lesion.31 Only a small proportion of subsequent seizures are truly “unprovoked,” and few studies distinguish seizures on the basis of timing. Seizures are most common (60–85%) with low-grade, slow-growing gliomas (90% with gangliogliomas), less common in anaplastic gliomas (54–69%), and least common in glioblastoma (29–49%).31,32⇓ Seizures occur in 29–41% of patients with meningiomas and in 35% of those with metastases.33 Location in the temporal lobe, primary motor and sensory cortices, and supplementary motor area appears to increase the risk for seizures.33 The issue of timing for neuroprotective therapy in tumors is of critical importance because many of the epileptogenic changes may have occurred before diagnosis and may therefore not be modifiable.

Neurosurgery.

Postoperative seizures are usually multifactorial, with some risk from the underlying condition necessitating surgery. The risk secondary to surgery alone is therefore difficult to determine. Early postoperative seizures (within 1 week of surgery) occur in about 10–15% of patients who undergo craniotomy for supratentorial brain tumors.34 Late or recurrent seizures are seen in 20–50%, most frequently in the setting of tumor recurrence and in those who had seizures before surgery.34

Neurodegenerative disorders, and Alzheimer’s disease (AD).

Alzheimer’s disease accounts for about 2% of all epilepsy.11 Seizures develop in approximately 10–17% of long-term survivors,11,35⇓ usually 5 to 10 years after disease onset. Specific risk factors for the development of seizures have not been determined, although many of the patients had disease onset at younger age.35 Because most of the patients have significant dementia when seizures begin, the impact of seizure prevention alone would be low in this group of patients.

Status epilepticus (SE).

Although animal models of SE are commonly used to study epileptogenesis, the impact of SE on the development of epilepsy in humans is difficult to determine. Many episodes of SE occur in the setting of brain insults that are, in themselves, risk factors for seizures. Other episodes of SE occur in patients who are already known to have epilepsy, further obscuring the causal relationship. In a retrospective cohort of patients with acute symptomatic seizures, a history of SE at the time of the initial seizure increased the 10-year cumulative incidence of later unprovoked seizures to 41%, compared to 13% in those without SE.36 A CNS etiology for the acute symptomatic SE also increased the risk of late seizures; those with anoxic insults had a relative risk of 18.8 and those with structural brain lesions 7.1, while metabolic SE was 3.6. These results suggest a synergistic effect of structural brain damage and SE for the development of late seizures. Alternatively, SE could be a marker for the severity of the underlying injury or could represent a biological substrate associated with a tendency to SE.

Febrile seizures.

Febrile seizures have a well-recognized association with epilepsy, but a causal link is uncertain. Febrile seizures may occur in patients with a genetic or structural susceptibility to seizures, rather than being the cause of late epilepsy. Simple febrile seizures (single, short duration, and generalized seizures in neurologically normal children) do not significantly increase the risk for subsequent epilepsy.37 Complex febrile seizures, on the other hand, are prolonged, often of focal onset, recurrent, occur in children with neurologic deficits, and do increase the risk for subsequent unprovoked seizures. In Rochester, Minnesota from 1935 to 1979, febrile seizures occurred in 2.3% of the population by age 5 years.37 The occurrence of febrile seizures resulted in a five-fold increased risk for epilepsy, which persisted to age 25 years. The cumulative incidence of epilepsy after febrile seizures was 6% by age 20, but the occurrence of epilepsy was not uniform throughout the population. Those with simple febrile seizures had a 5% risk for seizures by age 20, and those with complex features (abnormal neurologic exam, focal features, or duration greater than 10 minutes) had a greater risk. If one complex feature was present, the risk by age 20 was 6–8%; two features, 17%; and three features, 49%. Similarly, Nelson et al.38 followed patients with febrile seizures to age 7 years and found that 3% had an unprovoked seizure and 2% had epilepsy. The risk for development of epilepsy was again increased in patients with one or more complex features. The long and variable latent period in febrile seizures decreases the utility of this population in studies of antiepileptogenesis.

Future directions.

Existing studies of the development of epilepsy after brain insults have several limitations, and refinement of study design may significantly improve the ability to explore epileptogenesis. Acute symptomatic seizures and late/unprovoked seizures must be clearly distinguished and should be analyzed separately because they appear to have very different pathogeneses. Traumatic brain injury studies have defined early seizures, occurring in the first week, and late seizures, occurring after this period.17,39⇓ Prospective studies in other brain injuries should use similar criteria. Multivariate analysis of confounding factors, such as severity of injury and anatomic location, will help to better define the relationship between early and late seizures. If early seizures “cause” late seizures, drugs that retard kindling may be good candidates for antiepileptogenesis trials. Use of AEDs must be strictly monitored to avoid masking the true epileptogenic potential of the insult. Because many early seizures are subclinical, additional diagnostic methods, such as EEG monitoring, may be necessary to accurately diagnose seizures in acutely brain-injured patients.

Because no animal model completely represents the features of human epilepsy, longitudinal observational studies will be critical in determining which animal models are most applicable to human epileptogenesis. The mechanisms and timing of epileptogenesis after brain injuries are poorly defined. Identification and stratification of patient groups at high versus low risk for seizures after brain injury, or patients with and without seizures at presentation with a brain lesion such as a tumor, allow prospective evaluation of changes associated with the development of epilepsy. Important goals are to identify potential biomarkers of epileptogenesis and to select antiepileptogenic agents relevant to the mechanisms of a particular brain injury (e.g., iron chelator versus antioxidant versus NMDA antagonist). Serial or long-term EEG may be able to identify seizure precursors before the onset of clinically apparent seizures. Recent advances in neuroimaging, including MRI normative databases of the brain, will greatly enhance the localization of epileptogenic (or potentially epileptogenic) foci. Functional MRI and PET can then characterize local perturbations in activity presumably playing a role in epileptogenesis.

Finally, animal models of epileptogenesis show clear age-related differences in susceptibility to damage from seizures. Similar variability appears to occur in humans. For example, an initial precipitating injury such as head trauma or CNS infection at an age less than 5 years, is associated with a higher risk for mesial temporal sclerosis than are similar injuries occurring at older ages.40,41⇓ Further characterization of such age-related differences will be important in identification of markers of epileptogenesis and in design of neuroprotective studies.