Kaplan-Meier survival curves for patients with ischemic and hemorrhagic stroke showing the probability of remaining free of seizures after stroke. There was a significant difference (log-rank test) between the seizure event curves (P = .002).
Occurrence of the first seizure after stroke during set intervals. Most seizures occurred in the first 24 hours after stroke onset.
Disability of patients with and without seizures. The higher the Canadian Neurological Score, the less the neurological disability. A, Cortical cerebral infarction. Disability is greater in patients with seizure. B, Cortical cerebral hemorrhage. Disability is less in patients with seizure. Error bars represent SD.
Bladin CF, Alexandrov AV, Bellavance A, Bornstein N, Chambers B, Coté R, Lebrun L, Pirisi A, Norris JW, for the Seizures After Stroke Study Group. Seizures After StrokeA Prospective Multicenter Study. Arch Neurol. 2000;57(11):1617–1622. doi:10.1001/archneur.57.11.1617
Studies of seizures after stroke have largely been retrospective, with small patient numbers and limited statistical analysis. Much of the doctrine about seizures after stroke is not evidenced based.
To determine the incidence, outcome, and risk factors for seizures after stroke.
International, multicenter, prospective, analytic inception cohort study conducted for 34 months.
Patients and Setting
There were 2021 consecutive patients with acute stroke admitted to university teaching hospitals with established stroke units. After exclusion of 124 patients with previous epilepsy or without computed tomographic diagnosis, 1897 were available for analysis. Mean follow-up was 9 months.
Main Outcome Measures
Occurrence of 1 or more seizures after stroke, stroke disability, and death after stroke.
Seizures occurred in 168 (8.9%) of 1897 patients with stroke (28 [10.6%] of 265 with hemorrhagic and 140 [8.6%] of 1632 with ischemic stroke). On Kaplan-Meier survival analysis, patients with hemorrhagic stroke were at significantly greater risk of seizures (P = .002), with an almost 2-fold increase in risk of seizure after stroke (hazard ratio [HR], 1.85; 95% confidence interval [CI], 1.26-2.73; P = .002). On multivariate analysis, risk factors for seizures after ischemic stroke were cortical location of infarction (HR, 2.09; 95% CI, 1.19-3.68; P<.01) and stroke disability (HR, 2.10; 95% CI, 1.16-3.82; P<.02). The only risk factor for seizures after hemorrhagic stroke was cortical location (HR, 3.16; 95% CI, 1.35-7.40; P<.008). Recurrent seizures (epilepsy) occurred in 47 (2.5%) of 1897 patients. Late onset of the first seizure was an independent risk factor for epilepsy after ischemic stroke (HR, 12.37; 95% CI, 4.74-32.32; P<.001) but not after hemorrhagic stroke.
Seizures occur more commonly with hemorrhagic stroke than with ischemic stroke. Only a small minority later develop epilepsy. Patients with a disabling cortical infarct or a cortical hemorrhage are more likely to have seizures after stroke; those with late-onset seizures are at greater risk of epilepsy.
SEIZURES secondary to stroke have been recognized for many years and are considered by some authorities as a major cause of epilepsy in the elderly.1 Although the frequency of seizures after stroke is variously estimated at 4% to 10%,2- 9 many of these data were based on retrospective studies with variable follow-up, often without computed tomographic (CT) confirmation of the lesion, or on patient numbers so small that no reliable statistical analysis was possible. Often included were patients with arteriovenous malformations, brainstem strokes, subarachnoid hemorrhage, or a previous history of seizures or epilepsy. Previous assumptions such as seizures being more frequent in cerebral hemorrhage4,10 or cardioembolic stroke2,3,11 are not reliably evidence based.
Results of population-based studies6- 9 using multivariate analysis have indicated that risk factors for poststroke seizures include hemorrhagic stroke, cortical location of stroke, and severity of stroke. Other studies have proposed that the risk of epilepsy is greater in patients with late onset of the initial seizure after stroke,2,12,13 and early-onset seizures are considered a risk factor for stroke-related death.4,5,12,14,15
Black et al15 found seizures only in stroke patients with cerebral hemispheric lesions and established that because seizures occurred in less than 10% of patients, statistical evaluation would only be valid with prospective data in large numbers of patients. A multicenter study was therefore undertaken to investigate the frequency of seizures after stroke, their effect on stroke mortality, their effect on neurological and functional outcome, and the relation to underlying cerebral pathological lesions. Because a preliminary study15 found that no patients had seizures associated with brainstem stroke, in the present study only patients with hemispheric stroke were included. We believed that conclusions concerning the role of anticonvulsant drug therapy would be limited because, for ethical and methodological reasons, a placebo-controlled study would not be practical.
The Seizures After Stroke Study (SASS) was a prospective, multicenter study undertaken by the Stroke Research Unit, University of Toronto, Ontario, and involving university teaching hospitals in Canada, Australia, Israel, and Italy. Patients were recruited during 34 months. All consecutive patients with acute stroke admitted to these hospitals were enrolled except those with (1) brainstem and cerebellar stroke; (2) subarachnoid hemorrhage, arteriovenous vascular malformations, subdural hematoma, or inflammatory vascular disease; (3) transient ischemic attack, a history of previous seizures or epilepsy, or "pseudostroke," eg, strokelike symptoms due to metabolic dysfunction, brain tumor, migraine, or psychiatric disorders; and (4) early death precluding CT confirmation of diagnosis.
At each hospital all patients were evaluated by a neurologist, and seizures were classified according to the International League Against Epilepsy criteria.16 Epilepsy was defined as a condition characterized by recurrent seizures separated by more than 24 hours. As in previous studies,17 a cluster of seizures occurring in a single 24-hour period was considered a single seizure episode. "Early onset" refers to the first seizure occurring 2 weeks or less after stroke onset and "late onset" refers to the first seizure occurring more than 2 weeks after stroke onset. Patients underwent a standardized clinical neurological evaluation, with disability graded according to a modified Canadian Neurological Score (recorded as Canadian Neurological Score × 10)18 on days 1, 3, 7, and 14, and CT imaging on admission to the hospital and at approximately 7 to 10 days.
To determine the pathogenesis of ischemic stroke, all patients were investigated in a standardized fashion, including duplex ultrasound to assess carotid stenosis; the likelihood of cardioembolic stroke was graded as high or low using clinical and laboratory data.19 Lacunar infarction was defined according to clinically recognized lacunar syndromes, with or without CT evidence of a small deep infarct.20 For intracerebral hemorrhage, a previously validated scoring system was used to categorize the amount of blood in the ventricular and cisternal cerebrospinal fluid spaces.21 Volumes of CT lesions were measured using a computerized technique (Sigma Scan; Jandel Scientific, Corte Madera, Calif). An electroencephalogram was not a requirement for inclusion in the study but was performed if possible within the first week of admission.
The study protocol was approved by the research ethics committee (or equivalent) in each participating hospital. All data and CT scans were recorded in a standardized format and sent to the coordinating center (Stroke Research Unit, University of Toronto).
Follow-up of surviving patients was by telephone at approximately yearly intervals, or sooner if seizures were reported. When necessary, family relatives or the familyphysician, hospital, or nursing home was contacted. Patients identified as having symptoms suggestive of seizures attended neurology outpatient clinics for further evaluation. Functional outcome was assessed by the Modified Rankin Scale.
Univariate comparisons of characteristics between patients with and without seizures after stroke were performed using the χ2 test for categorical variables and the 2-sample t test for continuous variables.
Multivariate analysis was performed using a Cox proportional hazards model. Covariates in the Cox multivariate regression analysis of seizures after stroke were determined a priori based on review of the literature and results of univariate analysis. Variables in the Cox proportional hazards model for risk of seizures after ischemic stroke were age, presence or absence of hemorrhagic infarction, high or low risk of cardioembolism, size of infarct on CT, cortical location, and stroke disability. For the multivariate analysis for risk of recurrent seizures after ischemic stroke, the same set of variables were used but with addition of late-onset seizures. For the multivariate analysis for risk of death, the same variables were again used but with addition of early-onset seizures. For the purposes of multivariate analysis, infarct size on CT was entered as a categorical variable: no measurable lesion (0 cm3), small (<10 cm3), medium (10 to <100 cm3), or large (≥100 cm3). Stroke disability was also entered as a categorical variable based on the Canadian Neurological Score: mild disability (>90-115), moderate disability (>50-90), or severe disability (30-50).22
Covariates in the Cox multivariate regression analysis for risk of seizures after hemorrhagic stroke were age, grading of amount of cisternal and ventricular blood, hemorrhage location (lobar hemorrhage abutting onto cortex is "cortical" and deep subcortical hemorrhage is "subcortical"), size of hemorrhage on CT, and stroke disability. For the multivariate analysis for risk of recurrent seizures after hemorrhagic stroke, the same set of variables were used with addition of late-onset seizures. For the multivariate analysis for risk of death, the same variables were again used but with addition of early-onset seizures. Hemorrhage size was entered as a categorical variable: small (<20 cm3), medium (20 to <50 cm3), large (≥50 cm3), and very large (100-320 cm3). Stroke disability was also entered as a categorical variable based on the Canadian Neurological Score as for seizures after ischemic stroke.
Kaplan-Meier survival analysis was carried out for the cohorts with ischemic and hemorrhagic stroke analyzing survival free of first ever seizure during follow-up. Log-rank analysis was performed to compare the event curves for patients with seizures after ischemic stroke and seizures after hemorrhagic stroke.
Of 2021 patients with stroke enrolled in the study, 68 did not undergo CT and 56 had epilepsy before their stroke, leaving 1897 patients available for analysis. During the study, seizures occurred in 168 patients (8.9%), including 140 (8.6%) of 1632 with ischemic stroke and 28 (10.6%) of 265 with hemorrhagic stroke. The Kaplan-Meier survival analysis indicated a significantly greater probability of seizures occurring in patients with hemorrhagic stroke (P = .002), with a 1-year actuarial risk of seizures in stroke survivors of 20% in patients with hemorrhagic stroke and 14% in patients with ischemic stroke (Figure 1). Recurrent seizures (epilepsy) developed in 47 (2.5%) of 1897 patients with stroke overall and in 47 (28.0%) of 168 patients with seizure. Follow-up (mean, 9 months) was available for 1841 patients (97.0%).
Mean ± SD patient age was 72.0 ± 11.5 years, with no difference between seizure and nonseizure groups according to age, sex, hemisphere side, previous stroke, or other risk factors for stroke or seizures (Table 1).
Cortical infarction was present in the majority of patients with seizures after ischemic stroke (Table 1). Seizures were reported in 8 (2.6%) of 307 patients with deep lacunar infarction, although confounding factors made this association tenuous (see the "Comment" section). Infarct size on CT was significantly larger in patients with seizures than in those without seizures. Hemorrhagic infarcts were significantly more common in patients with seizures. However, there was no significant difference in the numbers of patients with and without seizures with a high risk of cardioembolic stroke (Table 1).
Forty percent of seizures after ischemic stroke occurred in the first 24 hours (Figure 2). Early-onset seizures (≤2 weeks) occurred in 4.8% (78/1632) and late-onset seizures (>2 weeks) occurred in 3.8% (62/1632) of patients with ischemic stroke. Recurrent seizures (epilepsy) occurred in 55% (34/62) of patients with late-onset seizures. Partial seizures (including simple partial and secondarily generalized seizures) accounted for 53% (74/140) of seizures after ischemic stroke.
Compared with ischemic stroke patients without seizures, those with seizures had a significantly worse neurological score during the hospital stay and Rankin score on long-term follow-up (Table 1). This was primarily due to cortical infarction because no significant differences in morbidity were detected between patients with and without seizures with subcortical infarction (Figure 3A). Mortality was greater in patients with seizure at 30 days and 1 year (Table 1).
In patients with seizures after hemorrhagic stroke, the parenchymal hemorrhage was predominantly cortical in location. Overall, there was no significant difference in hemorrhage size between the seizure and nonseizure groups and no significant difference in the amount of cisternal or ventricular blood present (Table 1).
Fifty-seven percent of seizures after hemorrhagic stroke occurred in the first 24 hours (Figure 2). Early-onset seizures (≤2 weeks) occurred in 7.9% (21/265) and late-onset seizures (>2 weeks) occurred in 2.6% (7/265) of patients with hemorrhagic stroke. Recurrent seizures (epilepsy) occurred in 100% (7/7) of patients with late-onset seizures after hemorrhagic stroke. Partial seizures (including simple partial and secondarily generalized seizures) accounted for 50% (14/28) of seizures.
Compared with patients with seizures after ischemic stroke, patients with seizures after hemorrhagic stroke had significantly less early disability than patients without seizure and also possibly a trend to less disability on follow-up Rankin score (Table 1). This was particularly evident in patients with cortical cerebral hemorrhage (Figure 3B). There were no significant differences in disability between patients with and without seizure with deep cerebral hemorrhage. There was no difference in mortality between patients with and without seizure at 30 days or 1 year (Table 1).
Compared with ischemic stroke, there was an almost 2-fold increased risk of seizures in patients with hemorrhagic stroke (hazard ratio, 1.85; 95% confidence interval, 1.26-2.73; P = .002). For the total patient cohort, those with late-onset seizures were at greater risk of developing epilepsy (hazard ratio, 23.77; 95% confidence interval, 12.89-43.81; P<.001).
On Cox proportional hazards analysis, risk factors for seizures after ischemic stroke were cortical location of infarction and stroke disability (Table 2). Neither hemorrhagic infarction nor high embolic risk of stroke were significant. The only risk factor for epilepsy (recurrent seizures) was late onset (>2 weeks) of the initial seizure after ischemic stroke (Table 2).
The only risk factor for seizures after hemorrhagic stroke was cortical location (Table 3). Neither size of intracerebral hemorrhage nor cerebrospinal fluid blood was independently associated with seizures. No independent risk factors were identified for the development of epilepsy (recurrent seizures) after hemorrhagic stroke (Table 3).
Results of the SASS indicate that although 8.9% of patients with stroke experienced seizures during the study, epilepsy was a sequela in only 2.5%. On Kaplan-Meier analysis, patients with hemorrhagic stroke were at significantly greater risk of seizures (20% per year) compared with patients with ischemic stroke (14% per year), particularly in the first few days after stroke onset; proportional hazards analysis identified hemorrhagic stroke as an independent risk factor for seizures. This was a prospective, hospital-based study enrolling all consecutive patients with stroke. Patients without a CT scan were excluded from the study but constituted only a small proportion of the final study number.
The frequency of epilepsy as a late sequela of stroke has been estimated previously at 3% to 10%,2 with a higher risk after late-onset than early-onset seizures.2,12,13 In SASS, epilepsy occurred in 2.5% of patients overall but was present in approximately half of those with late-onset seizures after ischemic stroke and in all patients with late-onset seizures after hemorrhagic stroke. After controlling for other clinical variables, late-onset seizures were identified as an independent risk factor (with a 12-fold increase) only for epilepsy after ischemic stroke. Many factors, including patient age, disability, and risk of adverse effects, need to be considered before administration of antiepileptic medications.
The occurrence of early and late poststroke seizures parallels that of posttraumatic epilepsy.23 In this paradigm, seizures in the early phase are thought to be the result of cellular biochemical dysfunction, and late-onset seizures are thought to be due to gliosis and the development of a meningocerebral cicatrix.23 The sequestration of hemosiderin in cortical neurons might also play a role in neurotraumatic seizures.24 A probable cause of early seizures after ischemic stroke is the extent of regional metabolic dysfunction and excitotoxic neurotransmitter release secondary to ischemic hypoxia. Ischemic brain tissue (the "ischemic penumbra") might contain electrically irritable tissue that provides a focus for seizure activity in patients with ischemic stroke.25 Late-onset seizures have a similar frequency in ischemic and hemorrhagic lesions, possibly suggesting a nonspecific epileptogenic effect of gliotic scarring.
Patients with seizures after ischemic stroke had larger cortical infarcts than those without seizures, although on multivariate analysis only cortical location and stroke disability (not infarct size or hemorrhagic transformation) were independent risk factors for seizures. Although results of previous studies2,3,7,11 have suggested that seizures are more common with cardioembolic cerebral infarction, recent studies4,14,26 have questioned this. In a National Institute of Neurological and Communicative Disorders and Stroke study27 of 1290 patients with cerebral infarction, cardioembolic stroke was differentiated by a history of systemic embolism, sudden clinical onset, and loss of consciousness but not epileptic seizures.
The relation of seizures to small subcortical infarcts is uncertain, with a reported seizure frequency of 0% to 23%.5,9,13,14,28,29 Seizures are considered by some as an exclusion for lacunar stroke.20 In SASS, 8 (2.6%) of 307 patients diagnosed as having lacunar stroke had seizures. Of these, 5 had normal results on early CT but were noted to have an electroencephalographic or single photon emission CT result suggestive of cortical dysfunction and 2 had very late onset of their first seizures during follow-up (possibly not related to stroke). Only 1 (0.7%) of 140 patients was clearly identified as having lacunae on CT imaging and poststroke seizures, but a delirium precluded further investigation.
Cerebral hemorrhages giving rise to seizures were predominantly cortical in location, although one third of seizures arose from deep cerebral hemorrhages, similar to previous studies.13,30,31 There was no significant difference in overall lesion size between patients with and without seizures. On multivariate analysis, only cortical location (not hemorrhage size, disability, or cerebrospinal fluid blood) was identified as a risk factor for seizures.
In contrast to patients with seizures after ischemic stroke, patients with seizures after cortical hemorrhagic stroke had less disability than those without seizures. The reasons for this are unclear but might reflect the greater overall initial disability of intracerebral hemorrhage and the inability to detect seizures in severely disabled or unconscious patients. Previous studies also reveal that large cerebral hemorrhages do not cause seizures, although the associated neurological deficit is more severe,32 and seizures are not associated with clinical deterioration in patients with cerebral hemorrhage.33
What causes a cerebral hemorrhage to produce seizures is unclear. Seizures occurring in subarachnoid hemorrhage relate to the amount of blood in the basal cisterns,21 but our data and those of others30 reveal that neither the amount of blood present in brain parenchyma (ie, lesion size) nor in the cisternal and ventricular spaces seemed to affect the occurrence of seizures. A combination of the sudden development of a space-occupying lesion with mass effect, focal ischemia, and blood products might possibly account for seizures in the early phase of hemorrhagic stroke. Other studies31 have shown no association between seizures and the presence of hydrocephalus, hemorrhage size, intracranial shift, Glasgow Coma Scale score, level of consciousness, or degree of neurological deficit.
Finally, we did not attempt to address issues of treatment of seizures. To conduct a natural history study (ie, no treatment) or even a placebo-controlled trial is difficult logistically, if not unethical, because many physicians will treat patients with antiepileptic drugs if seizures occur.
Accepted for publication May 9, 2000.
We thank the following personnel who assisted in the collection and processing of data for the SASS Group: Sunnybrook Health Science Centre, Toronto, Ontario: Philippa Johnston, MSc; Beverly Bowyer, RN; Lilliana Smurawska, MD; Natasha Alexandrova, MD; Alastair Dempster; and Marietta Medel, MD; Hôpital Charles LeMoyne, Montreal, Quebec: Louise Bergeron, RN (research assistant); Montreal General Hospital, Montreal: Lisa Wadup and Susan Joseph, RN; Hôpital St Luc, Montreal: Marie-Paule Desrochers, RN, BSc; Ichilov Hospital, Tel Aviv, Israel: Vadim Karepov, MD; Austin and Repatriation Medical Centre, Melbourne, Australia: Kim Sukhong, RN; Peter Brimage, FRACP; and Lichun Quang; and Istituto di Clinica Neurologica, Sassari, Italy: Marco Zuddas, MD; Maria A. Spanu, MD; Caterina Fiori, MD; Giacomo Murino, MD; Franco Bandiera, MD; Luisa Pes, MD; and Francesco Flumene MD.
The SASS Group investigators at each participating hospital are as follows: Sunnybrook Health Science Centre: Christopher F. Bladin, MD, FRACP; Andrei V. Alexandrov, MD; John W. Norris, MD; and Sandra Black, MD (Stroke Research Unit, University of Toronto [coordinating center]); Hôpital Charles LeMoyne: André Bellavance, MD, PhD (Sherbrooke University, Montreal); Ichilov Hospital: Natan Bornstein, MD (University of Tel Aviv); Austin and Repatriation Medical Centre: Brian Chambers, MD (University of Melbourne); Montreal General Hospital: Robert Coté, MD, and Arienne McKay, MD (McGill University, Montreal); Hôpital St Luc: Louise Lebrun, MD (University of Montreal); Istituto di Clinica Neurologica: Angelo Pirisi, MD (University of Sassari); Statistics: Marco Katic and John Szalai, PhD (University of Toronto); and LiChun Quang (Austin and Repatriation Medical Centre, University of Melbourne).
Reprints: Christopher F. Bladin, MD, FRACP, Department of Neuroscience, Box Hill Hospital, Nelson Road, Box Hill 3128, Melbourne, Australia.