Seizures in Acute Stroke: Predictors and Prognostic Significance
The Copenhagen Stroke Study
Background Despite the common occurrence of seizures during the early phase of stroke (ES), the effect of ES on prognosis is not known. We determined the relationships between ES and stroke outcome and identified predictors of ES.
Methods In this community-based study, we prospectively and consecutively studied 1197 patients with acute stroke. We determined the number and type of seizures, initial stroke severity, infarct size, mortality, and outcome in survivors. Stroke severity was measured on admission, weekly, and at discharge using the Scandinavian Stroke Scale (SSS). Multiple logistic and linear regression outcome analyses included relevant confounders and potential predictors, including age, gender, stroke severity on admission, atrial fibrillation, ischemic heart disease, diabetes, blood glucose level on admission, claudication, and hypertension.
Results Fifty patients (4.2%) had seizures within 14 days of the stroke. In the multivariate analyses, only initial stroke severity was related to ES; stroke type and lesion localization were not related. For each 10-point increase in stroke severity (SSS score), the relative risk of ES increased by a factor of 1.65 (95% confidence interval, 1.4 to 1.9) (P<.0001). ES did not influence the risk of death during hospital stay (P=.56). In survivors, ES was related to a better outcome, equivalent to an increased SSS score of 5.7 points (SE [b]=1.8; P=.002).
Conclusions The decisive factor of ES was initial stroke severity. ES per se was not related to mortality. Surprisingly, in survivors, ES predicted a better outcome. We explain this finding by a relatively larger ischemic penumbra in patients who have an ES after a stroke.
Seizures are common during the early phase after a stroke and have been reported to occur with a frequency of 2.5% to 5.7% within 14 days after a stroke.1 2 3 4 The effect of ES on prognosis is not known. If ES worsens prognosis, the prophylactic use of anticonvulsants in patients at risk is warranted.
Previous studies of ES were based on small numbers of selected patients, and neither multivariate analyses nor validated neurological scoring systems were used.1 2 3 4 5 6 7 8 To identify predictors of ES and to determine whether ES is directly related to mortality and neurological outcome, one must take into account relevant confounders and potential predictors such as age, gender, stroke severity, and other relevant risk factors; however, in previous studies, this was not done.
In the Copenhagen Stroke Study, we examined the relationship between ES and stroke outcome and predictors of ES in 1197 consecutive patients admitted with acute stroke.
Subjects and Methods
The study was prospective, consecutive, and community-based. Included were all patients with acute stroke admitted to the neurological department of Bispebjerg Hospital, Copenhagen, between September 1, 1991, and September 30, 1993.
The design of the Copenhagen Stroke Study is described in detail elsewhere.9 10 The 60-bed stroke unit within the neurological department serves all patients with stroke admitted to the hospital from a well-defined area of Copenhagen with 239 886 inhabitants, regardless of age, severity of stroke, or condition of the patient before the stroke. Hospital care is free, and the admission rate of patients with stroke in the Copenhagen area is 88%.11 Diagnostic procedures, treatment, and all stages of rehabilitation take place within the stroke unit. Patients are not discharged until the rehabilitation team decides that further in-hospital improvement is unlikely; referral to other departments or hospitals for further rehabilitation is therefore not required. A total of 1197 patients with acute stroke were admitted during the inclusion period. Two of 52 patients with ES had known epilepsy and were excluded; consequently, a total of 1195 acute stroke patients were included in the study.
Stroke was defined according to World Health Organization criteria, that is, rapidly developed clinical signs of focal disturbance of cerebral function lasting more than 24 hours or leading to death with no apparent cause other than vascular origin. Subarachnoid bleeding was not included.12
Seizures were classified according to the recommendations of the International League Against Epilepsy.13 ES was defined as a seizure within 14 days of the stroke. The time interval between stroke and witnessed epileptic seizures, the number and type of seizures, and anticonvulsant treatment were recorded. Electroencephalography was performed in 15 (30%) patients with ES. It was not the aim of the study to assess the value of electroencephalography in predicting the risk of developing seizures.
CT was performed with a Siemens Somatom DR scanner. Contrast medium was not given routinely. The time from onset of stroke to CT examination depended on the accessibility of the scanner, which varied during the study period. The median time from stroke onset to CT was 8 days. All scans were described by the study radiologist (H.O.R.), who was blinded to clinical data. Lesion size was determined by the largest diameter visible on the CT scan.
The SSS was used to assess stroke severity.14 15 The SSS evaluates level of consciousness; eye movement; power in the arms, hands, and legs; orientation; aphasia; facial paresis; and gait. The total score ranges from 0 to 58 (maximum) points. It was performed on admission, the day after admission, every week during the hospital stay, and at discharge. The weekly assessment was performed by the same neurologist (H.S.J.) in all patients.
In the multivariate regression analyses of the relationship between ES and risk factors, between ES and CT characteristics, between ES and mortality, and between ES and outcome in survivors, the following factors were included as independent variables: age, gender, atrial fibrillation (verified by the electrocardiogram obtained on admission), intermittent claudication, hypertension (considered present if the patient was under treatment with antihypertensive drugs at the time of admission or if hypertension was diagnosed during the hospital stay), diabetes mellitus (a history of diabetes or diabetes diagnosed during the hospital stay), blood glucose level on admission, ischemic heart disease, and initial stroke severity. Because initial stroke severity (SSS score on admission) may be influenced by seizures occurring before the initial rating, subanalyses of outcome were performed in patients with onset of seizures after assessment of initial stroke severity.
Statistical analyses were performed using the SPSS software package for Windows® 6.0.16 In univariate analyses, Student’s t test was used for continuous data, and the χ2 test was used for noncontinuous data. In the multivariate analyses, a logistic multiple regression model was used when the dependent variable was dichotomous, and a multiple linear regression model was used when the dependent variable was continuous. The level of significance was set at P<.05.
The study was approved by the Ethics Committee of Copenhagen (approval no. V. 100.2263/91).
The basic characteristics of all patients are listed in Table 1⇓. ES occurred in 50 patients (4.2%). The majority of the ES occurred within the first 72 hours (86%), particularly within the first 24 hours (66%). Twenty patients (40%) had a single seizure, and most seizures (34, or 68%) were simple partial seizures or partial seizures with 2° generalization. Eleven of the seizures (22%) were characterized as primary generalized tonic-clonic seizures. Twenty-three patients had onset of seizure after assessment of initial stroke severity, and 27 had onset of seizure before assessment of initial stroke severity. Forty-three of the patients (86%) were given anticonvulsants.
Patients with ES had more severe stroke (mean SSS score, 21; SD, 17) than patients without ES (mean SSS score, 37; SD, 17) (P<.0001). The relationship between stroke severity on admission (SSS score) and ES is shown in the Figure⇓. The occurrence of ES increased as a function of stroke severity from 1% in patients with an SSS score >45 to 11% in patients with an SSS score <15 (P<.00001). In the logistic multiple regression model, the risk of ES increased independently as a function of stroke severity. For each 10-point increase in stroke severity (SSS score), the relative risk of ES increased by a factor of 1.65 (95% CI, 1.4 to 1.9) (P<.0001).
The distribution of risk factors and results of univariate and multivariate analyses are given in Table 1⇑. In the univariate analyses, significant differences between patients with ES and those without ES were found for atrial fibrillation, diabetes, blood glucose level on admission, and initial stroke severity, whereas no difference was found for age, gender, claudication, ischemic heart disease, and hypertension. A multiple logistic regression analysis was performed to determine the independent influence of the factors analyzed. Only initial stroke severity significantly favored the occurrence of ES (P<.001). Factors included in the analysis but with no independent predictive power of ES were age, gender, atrial fibrillation, diabetes, blood glucose level on admission, claudication, ischemic heart disease, and hypertension.
Data obtained by CT in patients with and without ES are listed in Table 2⇓. A CT scan was obtained in 70% of the patients with ES and in 82% of the patients without ES. CT was performed in fewer patients with ES because these patients generally had more severe strokes and a higher rate of mortality; therefore, relatively more patients with ES died before CT could be performed or were unable to undergo examination because of their poor condition. Of 75 patients with ICH, 6 (8%) had ES, as did 29 (3%) of 900 patients with CT-verified ischemic stroke (P<.03). In the univariate analyses, the stroke lesion was larger in patients with ES than in patients without ES, the relative frequency of ICH in patients with ES (17%) was higher than in patients without ES (7%), the occurrence of silent infarcts was lower in patients with ES than in patients without ES, and cortical involvement was more frequent in patients with ES (73%) than in patients without ES (46%). In the multiple logistic regression analysis with ES as the dependent variable and initial stroke severity (SSS score), gender, age, ischemic heart disease, atrial fibrillation, ICH, silent infarction, and cortical involvement as independent variables, neither the occurrence of ICH, silent infarction, nor cortical involvement significantly influenced the risk of ES. The decisive factor was initial stroke severity.
Data concerning outcome in patients with and without ES are listed in Table 3⇓. In the univariate analysis, the mortality of patients with ES was 50%, compared with 20% in patients without ES (P<.0001). In the multiple logistic regression analysis with mortality as the dependent variable, ES did not significantly influence the risk of death during hospital stay (P=.56). This was also the case in the subanalysis including only patients with onset of seizures after assessment of initial stroke severity. The decisive factor was initial stroke severity. The number of patients was too low to perform a multivariate subgroup analysis of outcome with regard to type and severity of seizures.
In survivors, the occurrence of ES was related to a better outcome, equivalent to an increased SSS score of 5.7 points (SE [b]=1.8; P=.002). In the subanalysis of survivors with onset of seizures after assessment of initial stroke severity, the occurrence of ES was also related to a better outcome, equivalent to an increased SSS score of 6.1 points (SE [b]=2.7, P=.02).
ES occurred with a frequency, time pattern, and seizure subtype distribution consistent with those of previous studies.1 2 3 7 In the univariate analyses, there was no relationship between ES and age, gender, ischemic heart disease, claudication, or hypertension, but relationships were found between ES and atrial fibrillation, diabetes, blood glucose level on admission, and initial stroke severity. However, in the multivariate analysis, only the relationship with initial stroke severity was statistically significant, making initial stroke severity the most important predictor of ES.
Some,5 7 8 but not all,1 previous studies have indicated a relationship between infarct size and ES. Results obtained in this study confirm the existence of such a relationship, consistent with the observed higher initial stroke severity in patients with ES. Some previous studies,1 5 but not others,3 4 6 have indicated relationships between ES and ICH. We found a higher frequency of ES in patients with ICH than in those with cerebral infarcts in the univariate analysis, but in the multivariate analysis, initial stroke severity was the sole predictor of ES. The apparent difference observed in the univariate analysis exclusively reflects a higher initial stroke severity, which has been shown to exist in ICH17 ; thus, the subtype of stroke per se does not influence the risk of ES.
A relationship between ES and cortical involvement has been reported in some studies1 5 but not in others.6 7 We found a highly significant association between cortical involvement and occurrence of ES in the univariate analysis. However, in the multivariate analysis, initial stroke severity was the sole predictor of ES. It should be stressed that this finding by no means rules out the cortex as the source of ES. The area of reduced flow (ie, the infarct and the surrounding penumbra) determines the severity of stroke in the acute state. The ischemic penumbra is not visualized by CT. If still-viable cortex is included in the penumbra, this would explain why stroke severity was the sole predictor of ES in this study and why cortical involvement (demonstrated on CT scans) dropped out as a predictor in the multivariate analysis. If we had performed single-photon emission CT or positron emission tomography instead of CT, cortical involvement would have been detected in many more patients with ES and, together with stroke severity, may have turned out to be a predictor of ES.
In the univariate analysis, the mortality rate of stroke patients with ES was 2.5-fold higher than that of patients without ES, but in the multivariate analysis, the high mortality rate in patients with ES was not related to seizures but explained exclusively by initial stroke severity. The high mortality rate of patients with ES is in striking contrast to results of the only other large prospective study, in which the mortality rate of patients with ES was 20%.1 Although this mortality rate was not significantly different from the 17% mortality rate found in patients without ES, 26% of the patients in the study had transient ischemic attacks or subarachnoid hemorrhage.1 A trend toward higher mortality in patients with ES has been reported in one other study.4
A key question is whether ES per se worsens prognosis. Assessment of outcome in survivors with ES has been attempted in only one study; no association with ES was found, but no validated score system was used, and no discrimination between ES occurring before and after assessment of initial stroke severity was made.1 A priori it is likely that ES worsens prognosis, because seizures accelerate cerebral glucose metabolism several-fold and contribute to increased levels of deleterious lactate. Because the majority of seizures occurred within 24 hours after stroke in the temporal window, where a penumbra of potentially salvageable brain tissue is believed to exist,18 19 it may be speculated that ES per se led to the observed larger infarct measured by CT. Although patients with ES had more severe strokes, their outcome was not significantly different from that of patients without ES in the univariate analysis. Surprisingly, however, surviving patients with ES, including the subgroup of patients with onset of seizure after assessment of initial stroke severity, turned out to have a better outcome when initial stroke severity was taken into account in the multivariate analysis. It cannot be concluded that the seizures per se improve the prognosis in survivors. ES could still be harmful but coexist with a higher frequency in strokes with more nonfunctional but viable brain tissue. We propose that the better outcome is explained by a larger ischemic penumbra, which was recently shown to exist in patients with stroke with ES.20 A large penumbra is related to a possible good outcome but also to enhanced risk of seizures in metabolically disturbed tissue because neurons are still able to discharge. Initial stroke severity (SSS score) is determined by the amount of dead tissue in the core of the infarct as well as the amount of nonfunctional but potentially viable tissue in the penumbra. If neurons of the penumbra survive, a better outcome (gain) will be positively correlated with the relative size of the penumbra. Several known pathophysiological processes of the penumbra may enhance the risk of seizure development. In contrast to neurons in the core zone of the infarct, neurons in the penumbra are still alive and able to discharge. Epileptogenic processes such as enhanced release of excitotoxic glutamate, ionic imbalances, breakdown of membrane phospholipids, and release of free fatty acids all characterize the penumbra.
Selected Abbreviations and Acronyms
|ES||=||occurrence of seizure during the early phase after a stroke|
|SSS||=||Scandinavian Stroke Scale|
Supported by grants from the Danish Health Foundation, the Danish Heart Foundation, and the Ebba Celinders Foundation.
Reprint requests to Jakob Reith, MD, Department of Neurology, Bispebjerg Hospital, Bispebjerg Bakke 23, DK-2400 Copenhagen, Denmark.
- Received April 4, 1997.
- Revision received May 15, 1997.
- Accepted May 15, 1997.
- Copyright © 1997 by American Heart Association
Black SE, Norris JW, Hachinski VC. Post-stroke seizures. Stroke. 1983;14:134. Abstract.
Shinton RA, Gill JS, Melnick SC, Gupta AK, Beevers DG. The frequency, characteristics and prognosis of epileptic seizures at the onset of stroke. J Neurol Neurosurg Psychiatry. 1988;51:273-276.
Gupta SR, Naheedy M, Elias D, Rubino FA. Postinfarction seizures: a clinical study. Stroke. 1988;19:1477-1481.
Olsen TS, Høgenhaven H, Thage O. Epilepsy after stroke. Neurology. 1987;37:1209-1211.
Jørgensen HS, Nakayama H, Raaschou HO, Gam J, Olsen TS. Silent infarction in acute stroke patients: prevalence, localization, risk factors, and clinical significance. The Copenhagen Stroke Study. Stroke. 1994;25:97-104.
Jørgensen HS, Plesner AM, Hubbe P, Larsen K. Marked increase of stroke incidence in men between 1972 and 1990 in Frederiksberg, Denmark. Stroke. 1992;23:1701-1704.
WHO Task Force on Stroke and Other Cerebrovascular Disorders. Report of the WHO Task Force on Stroke and Other Cerebrovascular Disorders: stroke–1989. Recommendations on stroke prevention, diagnosis, and therapy. Stroke. 1989;20:1407-1431.
Commission on Classification and Terminology of the International League Against Epilepsy. Proposal for revised clinical and electroencephalographic classification of epileptic seizures. Epilepsia. 1981;2:489-501.
Scandinavian Stroke Study Group. Multicenter trial of hemodilution in ischemic stroke: background and study protocol. Stroke. 1985;16:885-890.
SPSS. Statistical Package for Social Sciences, Windows® version, release 5.0. Chicago, Ill: SPSS; 1992.
Jørgensen HS, Nakayama H, Raaschou HO, Olsen TS. Intracerebral hemorrhage versus infarction: stroke severity, risk factors, and prognosis. Ann Neurol. 1995,38:45-50.
Olsen TS, Larsen B, Herning M, Skriver EB, Lassen NA. Blood flow and vascular reactivity in collaterally perfused brain tissue: evidence of an ischemic penumbra in patients with acute stroke. Stroke. 1983;14:332-342.
Wise RJS, Bernardi S, Frackowiak RSJ, Legg NJ, Jones T. Serial observations on the pathophysiology of acute stroke. Brain. 1983;106:197-222.
Bladin CF, Alexandrov AV, Norris JW. Epileptic seizures after stroke. Stroke. 1996;27:179. Abstract.