Epileptic seizures after a first stroke: the Oxfordshire community stroke projectBMJ 1997; 315 doi: http://dx.doi.org/10.1136/bmj.315.7122.1582 (Published 13 December 1997) Cite this as: BMJ 1997;315:1582
- John Burn, consultant in rehabilitation medicinea (, )
- Martin Dennis, senior lecturer in stroke medicineb,
- John Bamford, consultant neurologistc,
- Peter Sandercock, reader in medical neurologyb,
- Derick Wade, consultant neurologistd,
- Charles Warlow, professor of medical neurologyb
- a Rehabilitation Research Unit, Southampton General Hospital, Southampton SO9 4XY
- b Department of Clinical Neurosciences, Western General Hospital, Edinburgh EH4 2XU
- c Department of Neurology, St James's University Hospital, Leeds LS9 7TF
- d Rivermead Rehabilitation Centre, Oxford OX1 4XD
- Correspondence to: Dr Burn
- Accepted 8 July 1997
Objective: To describe the immediate and long term risk of epileptic seizures after a first ever stroke.
Design: Cohort study following up stroke survivors for 2 to 6.5 years; comparison with age specific incidence rates of epileptic seizures in the general population.
Setting: Community based stroke register.
Subjects: 675 patients with a first stroke, followed up for a minimum of 2 years.
Main outcome measures: Occurrence of single and recurrent seizures.
Results: 52 patients had one or more post stroke seizures; in 25 the seizures were recurrent. The 5 year actuarial risk of a post stroke seizure in survivors (excluding 19 patients with a history of epilepsy and 3 patients in whom the seizure occurred shortly before death from another cause) was 11.5% (95% confidence interval 4.8% to 18.2%). The relative risk of seizures, in comparison with the general population, was estimated at 35.2 in the first year after stroke and 19.0 in year 2. The risk of seizures was increased in survivors of subarachnoid and intracerebral haemorrhage (hazard ratio for intracranial haemorrhage v cerebral infarction 10.2 (3.7 to 27.9)). The risk of seizures after ischaemic stroke was substantial only in patients presenting with severe strokes due to total anterior circulation infarction. Only 9 of 295 patients (3%) independent one month after stroke suffered a seizure between 1 month and 5 years (actuarial risk 4.2% (0.1% to 8.3%)).
Conclusion: Stroke patients have about an 11.5% risk of single or recurrent seizures in the first 5 years after a stroke. Patients with more severe strokes or haemorrhagic strokes are at higher risk.
Cerebrovascular disease is an important cause of epilepsy but the risk of seizures after stroke may be overestimated by hospital based studies
From a community based register, patients with a first ever stroke had a 2% risk of having a seizure at stroke onset and an 11% risk of having a later seizure in the first 5 years of follow up
Patients with intracerebral and subarachnoid haemorrhage were at higher risk of seizures after stroke
Survivors who were independent at 1 month were at very low risk of future seizures
Cerebrovascular disease is an important cause of epilepsy,1 particularly in elderly people.2 When seizures complicate a clinical stroke they have a devastating effect on morale and further impair an already compromised quality of life. Precise estimates of the risk of developing epilepsy would be helpful not only to patients but also to those who give advice on returning to work or driving. Data from hospital based studies may give falsely high estimates of risk since patients with severe strokes, or strokes presenting with seizures, may be more likely to be admitted to hospital.3 In this study we had three aims: to describe the immediate and long term risk of epileptic seizures after a first stroke among patients in a community stroke register; to analyse the occurrence of seizures in relation to the pathological and clinical subtype of first stroke; and to compare the risk of a first epileptic seizure in these stroke patients with the risk in the general population.
The Oxfordshire community stroke project registered 675 patients with a first stroke, 357 (53%) women and 318 (47%) men, from a study population of about 105 000 over 4 years. Patients of all ages were included: 163 (24%) were under 65 years, 195 (29%) 65–74 years, 228 (34%) 75–84 years, and 89 (13%) over 84 years (mean age 72.2 years). Patients were recruited between 1981 and 1986 and were followed up until 1988, when the project finished (minimum follow up 2 years). The methods used to identify and assess cases have been described elsewhere.4 5 A total of 545 patients (81%) had cerebral infarction, 66 (10%) primary intracerebral haemorrhage, and 33 (5%) subarachnoid haemorrhage. In 31 (4%) cases the type of stroke could not be determined.
Ten patients had intracranial operations for aneurysm, and five of these patients received prophylactic phenytoin for 18 months to 2 years. Prophylactic anticonvulsant treatment was not offered to any other patients. One patient with primary intracerebral haemorrhage (not complicated by seizures) had intracranial surgery.
All but two of the 545 patients with cerebral infarction could be classified into subtypes according to their clinical presentation6: 92 patients had total anterior circulation infarction (cortical and subcortical ischaemia likely to be due to occlusion of the main stem of the middle cerebral artery or of the internal carotid artery); 185 patients had partial anterior circulation infarction (predominantly cortical ischaemia likely to be due to occlusion of a branch of the middle cerebral or anterior cerebral artery); 137 patients had lacunar infarction (a lacunar syndrome likely to be due to intracerebral small vessel occlusion7); and 129 patients had posterior circulation infarction (ischaemia in the brainstem, cerebellum or occipital cortex likely to be due to either small or large vessel occlusion).
Survivors were interviewed by a study nurse at 1 month, 6 months, 12 months, and then annually around the anniversary of their first stroke. At each follow up the nurse asked the patient and caregivers specifically about whether any possible seizure had occurred. As a check the nurse also reviewed the records kept by the primary healthcare team. Patients with suspected seizures were reassessed clinically by the study neurologist. Long term survivors were visited at the close of the study by one of the study neurologists (JPSB), and all available information about possible seizures was reviewed again with the patient. The hospital and primary care records of patients who died before this final visit were reviewed by one of the study neurologists (JPSB), and data relating to possible seizures were discussed at regular consensus meetings of the study team. Generalised seizures were diagnosed with reference to statements of witnesses. Focal seizures were distinguished clinically from clonus, spasms, and recurrent strokes.8 9 Electroencephalograms were not routinely performed. Each visit included an assessment of functional ability using the Oxford handicap scale10; in this report patients in grades 0–2 were described as independent and patients in grades 3–5 as dependent.
The Oxfordshire community stroke project followed up every patient for 2 years or until death. Patients recruited earlier in the study were followed for longer, and no patient was lost to follow up. Seventeen patients died on day 1, 129 by 30 days, 208 by 1 year, and 254 by 2 years. A total of 421 patients were followed up at 2 years, 274 were seen at 3 years, 182 at 4 years, and 92 at 5 years; 324 patients survived to a final examination by a study neurologist between August 1987 and November 1988.
Classification of seizures
Onset seizures were defined as occurring within 24 hours of the onset of stroke. This definition (see Appendix 1) is based on that used in a population based study of acute seizures after head injury.11 Seizures occurring later were classified as post stroke seizures, except in two cases where evolution of stroke was prolonged.
Comparison with the general population
The incidence of post stroke seizures was compared with the estimated rate of epileptic seizures in the general population, using data from the VAMP research bank, a large British primary care database.12 13 Eighty two practices had sent in data on the medical and prescribing history of 369 819 patients. The data were checked for completeness and internal consistency,13 and practices supplying unreliable data were excluded.
Actuarial analysis was used for cohort data involving the follow up of patients for varying lengths of time. Kaplan-Meier survival curves were constructed in which patients who died or were no longer being followed up (after 2 years) were censored. The occurrence of a post stroke seizure qualified as an end point, and these were graphically represented as a cumulative seizure rate. Different pathological types, age stratifications, and clinical subtypes of stroke were compared by using the log rank method.14 The incidence of epileptic seizures was compared between stroke survivors and the general population and expressed as a ratio of the observed to the expected frequency with confidence intervals calculated from the Poisson distribution.15 Calculations were made for year 1 and year 2 separately; there were insufficient events to continue this comparison in years 3-5. Confidence intervals of small samples were calculated by using an exact test.15
Seizures preceding stroke
Nineteen of the 675 patients (3%) registered with the Oxfordshire community stroke project gave a history of one or more seizures before their first stroke or had documentation of past seizures in their medical record. In one patient the seizures were secondary to a craniotomy for a benign tumour; in one they occurred in the context of eclampsia; and in two they were related to alcohol abuse. The remaining seizures were considered idiopathic, and some may have been caused by otherwise asymptomatic cerebrovascular disease. Eleven of the 19 patients had had a seizure in the year before the stroke: in seven this was a first seizure, and four of these seven patients had partial motor seizures in the same limb as was subsequently affected by the stroke.
Seizures at the onset of the first stroke
Fourteen patients (2%) had an onset seizure, none of whom had a previous history of seizures. In seven the seizure was generalised, in six it was a simple partial seizure, and in one it was a complex partial seizure.
The occurrence of an onset seizure was not, in this cohort, associated with a worse outcome. The 30 day case fatality rate of patients with onset seizures was not raised. Two patients (14%) with onset seizures died compared with 127 of the 661 without onset seizures (19.2%, 95% confidence interval 16.2% to 22.2%). There was also no suggestion that patients with onset seizures had a worse functional prognosis. Three of ten previously independent patients with onset seizures were dependent at one month compared with 179 of those without onset seizures (38.3%, 33.9% to 42.7%).
The occurrence of an onset seizure was, however, associated with an increased risk of having further seizures. Five (36%) of the 14 patients with onset seizures went on to have one or more post stroke seizures compared with 43 (6.9%) of 625 patients who survived the first day without a history of pre-stroke or onset seizures (odds ratio 7.52, 2.46 to 22.98).
Patients with subarachnoid haemorrhage had an excess of onset seizures but this was not statistically significant. Four (4%) of the 99 patients with intracranial haemorrhage had an onset seizure compared with 10 (1.8%) of the 545 patients with cerebral infarction (odds ratio 2.25, 0.69 to 7.33) (Table 1).
Post stroke seizures
Fifty two (8%) of 658 patients who survived the first day had one or more post stroke seizures. Three patients had their first seizure only a few hours before death as part of the terminal illness and one patient had had seizures before the stroke. Most analyses were restricted to the remaining 48 patients. Twenty three patients had only a single post stroke seizure; 25 patients had recurrent post stroke seizures as defined in the appendix (Table 2), but these generally occurred infrequently; only eight patients had more than one seizure a month over a period of two months or more. Only two of the five patients who had previously had an onset seizure went on to develop recurrent seizures.
Twenty four (50%) of the 48 patients were treated with anticonvulsant drugs; six of these had had only one post stroke seizure. Forty (83%) patients had at least one generalised seizure. Partial seizures occurring alone were less common than at the onset of the stroke.
Risk of post stroke seizure
A Kaplan-Meier survival curve was constructed in which the occurrence of a first post stroke seizure was recorded as an end point and deaths were censored (table 3, fig 1). Patients with a history of epilepsy before the stroke (n=19) or who died in the first day after the stroke (n=17) were excluded from these analyses. The probability of having a post stroke seizure was 5.7% (3.5% to 7.9%) within the first year and 11.5% (4.8% to 18.2%) within 5 years. The incremental risk averaged about 1.5% a year after the first 12 months. Five of the 10 patients who had a first seizure more than 2 years after the stroke had had another stroke more recently, and two had taken prophylactic phenytoin for two years after a subarachnoid haemorrhage.
The probability of a post stroke seizure was higher among patients aged over 84 years (observed/expected 2.0) but there was no significant trend on log rank analysis for an increasing risk with age (χ2=0.91, df=1, P>0.25). Subsequent analyses were not adjusted for age.
Patients who had had a haemorrhagic first stroke were more at risk of post stroke seizures than patients who had cerebral infarction (hazard ratio 10.2, 3.7 to 27.9) (Table 3); fig 2. However, the lower risk after cerebral infarction concealed a significant heterogeneity between different subtypes (χ2=34.0, df=3, P<0.001) (table 2); fig 3.
Survivors of total anterior circulation infarction had a markedly increased risk of post stroke seizures when compared with the survivors of other subtypes of cerebral infarction.
The good prognosis for survivors of partial anterior circulation infarction, lacunar infarction, and posterior circulation infarction was reflected in the low probability of future post stroke seizures among independent stroke survivors (those with Oxford handicap scale score 0-2). A further Kaplan-Meier curve restricted to patients alive and independent at 1 month revealed an incremental risk of suffering a post stroke seizure of about 0.9% a year and an actuarial risk at 5 years of 4.2% (0.1% to 8.3%). If only survivors independent at 6 months were included the 5 year actuarial risk was 2.7% (0 to 6.0%).
Comparison with the general population
Forty eight patients had at least one post stroke seizure recorded in a total follow up period (for patients without a history of epilepsy) of 1641 patient years—an incidence of 29.3 patients developing post stroke seizures per 1000 patient years, or 23.5 per 1000 patient years when patients with intracerebral infarction, subarachnoid haemorrhage, or unknown pathology were excluded. This incidence was compared year by year for the first 2 years after stroke with the age specific incidence in the general population, as recorded by the VAMP research data bank over the 12 months between 1 February 1989 and 31 January 1990. Stroke survivors had a 35-fold increased risk of seizures in the first year (Table 4). The increase in risk was particularly evident for patients under 65 years.
An earlier, community based study of seizures after stroke reported that 9.5% of 518 survivors suffered a seizure 2–5 years after stroke, but no mention was made of subjects who were not followed up or who had died before the assessment.16 Such problems can be addressed by an actuarial analysis which censors patients who die or who are lost to follow up. This method of analysis has been used in hospital based studies of epilepsy after stroke in New York17 and Sweden18 and in a population based study of seizures after cerebral infarction in Rochester, Minnesota.19 The 5% risk of developing recurrent seizures by 5 years after the stroke found in the Swedish study18 is similar to the risk observed in the Oxfordshire community stroke project (5.4%, 0.3% to 10.5%). The New York study reported a 19% risk of single or recurrent seizures by six years after an ischaemic stroke17; much higher than the risk we observed after cerebral infarction. This study, reported only in abstract, was retrospective and patients with post stroke seizures may have been preferentially followed up. An admission bias towards more patients with severe strokes20 may also have raised the rate of seizures.
Seizures before stroke
The identification of seizures occurring shortly before the stroke offers support for the concept of vascular precursor epilepsy.21 Eleven (2%) patients in the Oxfordshire study had a seizure in the year before the stroke. This was three times the number expected in the general population.22 A case-control study has shown an eightfold increased risk of epilepsy preceding stroke but was confounded by admission bias and the use of elective surgical admissions as controls,23 24 and a study of patients investigated with computed tomography at the time of their first clinical stroke has confirmed evidence of preceding asymptomatic cerebral infarctions in 11% of patients; half of these involved the cerebral cortex.25 Prestroke seizures could have arisen from such areas.
Seizures at onset of stroke
The Oxfordshire community stroke project recorded a lower rate of onset seizures than other studies which used a similar definition but in which a selection bias may have raised the rate.26 27 28 These studies, which found rates in excess of 4%, were restricted to patients admitted to hospital, and the much quoted study by Aring and Merritt included only cases coming to necropsy.28 This overrepresentation of fatal cases would have led to an excess of haemorrhagic strokes that were both epileptogenic and fatal.
A more representative hospital series of 1000 consecutive patients admitted to an Australian hospital with acute stroke identified 4.4% of patients with onset seizures and 3.5% with seizures after cerebral infarction.29 The incidence among patients in the Oxfordshire study was lower, but the Australian series used an extended definition of onset seizures, including events that occurred up to 2 weeks from stroke onset. With the same definition, four more Oxfordshire patients with ischaemic stroke would be included and the incidence of onset seizures after cerebral infarction would rise to 2.7% (1.5% to 3.9%), similar to the Australian result.
In the population based study from Rochester 4.8% of patients had an onset seizure within 24 hours of cerebral infarction,19 a higher rate than in the Oxfordshire study. The Rochester series may have included patients with haemorrhagic lesions as these subjects had sustained their first stroke between 1960 and 1969, before the use of computed tomography scanning. A more recent series of 1195 hospital patients with both ischaemic and haemorrhagic stroke, which used the same two week definition of onset seizures, reported an incidence of 4.2%.30 The occurrence of onset seizures in this study was associated, in a multiple regression analysis, with severity of the initial stroke, but the occurrence of an onset seizure did not in this study, or in the Oxfordshire study, necessarily indicate a worse outcome.
Onset seizures after stroke had features in common with acute seizures occurring after traumatic brain injury: the frequency of seizures occurring within 24 hours of the event was similar11 31; onset seizures were more often partial31 32 33; and the occurrence of an onset seizure increased the risk of later seizures.31 These similarities suggest that onset seizures after stroke and traumatic brain injury may have a common pathogenesis.34 35 The risk of later seizures is, if anything, higher after stroke,19 but such seizures are less likely to recur.31
Post stroke seizures
The risk of post stroke seizures was lower after cerebral infarction than after primary intracerebral haemorrhage or subarachnoid haemorrhage except among patients surviving total anterior circulation infarction, who had a 34% (12% to 57%) risk at 2 years. This estimated risk has broad confidence intervals because only 56 of these patients survived beyond the first 30 days, but it is significantly higher than after other subtypes of cerebral infarction. Post stroke seizures have previously been associated with severe paralytic stroke,36 particularly when the cerebral cortex has been involved.29 The increased incidence after total anterior circulation infarction may reflect the extensive damage frequently sustained to the frontal and temporal cortex, the most epileptogenic areas of the brain.37
Some patients in the Oxfordshire study developed post stroke seizures without any clinical or computed tomography evidence of cortical damage. Five patients developed post stroke seizures after lacunar infarction. One of these patients had a history of alcohol misuse and another had a dementing illness with evidence on postmortem examination of previous asymptomatic cerebral infarction. In the remaining three patients no other cause was found apart from the lacunar stroke. Other reports of seizures complicating basal ganglionic haemorrhage38 39 and lacunar infarction40 suggest that the association between post stroke seizures and cortical damage may not to be as exclusive as previously thought.36
The relative and cumulative risks of seizures after cerebral infarction were greater in the Oxfordshire community stroke project than in the one comparable study from Rochester,19 but both these studies showed a significant excess risk in the first year after stroke and a similar reduction in relative risk with increasing age. The differences may be due to chance alone (the Rochester results are all within the 95% confidence intervals of the estimates from the Oxfordshire study), but it is possible that the Oxfordshire study detected seizures that did not come to medical attention and so would not be included in the Rochester records.
Although the Oxfordshire study recorded an 11.5% risk of having a post stroke seizure by 5 years, a much higher risk than in the general population, almost half of these patients had only a single seizure, and of the 25 who had recurrent seizures only eight were having them frequently. It is unusual for epilepsy to be a major problem among stroke survivors,29 32 and the risk of seizures in this study was significant only after a haemorrhagic or severe ischaemic stroke. Stroke survivors who were functionally independent at 1 or 6 months after the stroke had a very low risk of future seizures. The results suggest a cautious approach to instituting anticonvulsant treatment and support the current British licensing regulations which allow stroke patients who are functionally competent to return to driving after 30 days.
We gratefully acknowledge support from the Medical Research Council (UK) and the Stroke Association. We also thank Lesley Jones, computer programmer, for her work in data management and statistical analysis, and the study nurses Sue Price, Liz Mogridge, and Claire Clifford. The project would not have been possible without the cooperation of the following general practices (liaison partners): Beaumont Street, Oxford (Dr A MacPherson); East Oxford Health Centre (Dr D Leggate); Thame (Dr A Marcus); Berinsfield (Dr M Agass); Stert Street, Abingdon (Dr N Crossley); Marcham Road, Abingdon (Dr R Pinches); Malthouse, Abingdon (Dr D Otterburn); Kidlington (Dr S Street); Wantage (Dr V Drury); Deddington (Dr H O'Donell). We also thank Mr Jim Slattery for statistical advice.
Funding: MRC and the Stroke Association.
Conflict of interest: None.
A single seizure includes a single episode of status epilepticus or multiple seizures with less than 2 weeks between events. Seizures which were “provoked” by a stroke recurrence, occurring at the onset of a new cerebrovascular event, were included as post stroke seizures but were not included in the definition of recurrent seizures.
Recurrent seizures—two or more unprovoked seizures with at least two weeks between events.29
Prestroke seizure—single or recurrent seizures occurring before the day of the first stroke.
Onset seizures—seizures occurring within 24 hours of stroke onset or, in two cases, at three and four days during the evolution of a progressive stroke while it was still getting worse.
Post stroke seizures—seizures occuring more than 24 hours after stroke onset or after a stable deficit had been established.