Incidence and Associations of Poststroke Epilepsy
The Prospective South London Stroke Register
Background and Purpose—To describe the epidemiology and associations of poststroke epilepsy (PSE) because there is limited evidence to inform clinicians and guide future research.
Methods—Data were collected from the population-based South London Stroke Register of first strokes in a multiethnic inner-city population with a maximum follow-up of 12 years. Self-completed forms and interviews notified study organizers of epilepsy diagnosis. Kaplan–Meier methods and Cox models were used to assess associations with sociodemographic factors, clinical features, stroke subtype, and severity markers.
Results—Three thousand three-hundred ten patients with no history of epilepsy presented with first stroke between 1995 and 2007, with a mean follow-up of 3.8 years. Two-hundred thirteen subjects (6.4%) had development of PSE. PSE incidence at 3 months and 1, 5, and 10 years were estimated at 1.5%, 3.5%, 9.0%, and 12.4%, respectively. Sex, ethnicity, and socioeconomic status were not associations, but markers of cortical location, including dysphasia, visual neglect, and field defect, along with stroke severity indices at presentation, including low Glasgow Coma Scale, incontinence, or poor function on Barthel Index, were associated with PSE on univariate analysis. Young age was independently associated with PSE, affecting 10.7% of patients aged <65 years and 1.6% >85 years (P≤0.001) on 10-year estimates. Independent predictors of PSE also included visual neglect, dysphasia, and stroke subtype, particularly total anterior circulation infarcts. Dysarthria was associated with reduced incidence.
Conclusions—PSE is common, with risk continuing to increase outside the acute phase. Young age, cortical location, larger lesions, and hemorrhagic lesions are independent predictors.
Cerebrovascular disease is the most commonly identified cause of acute symptomatic seizures and secondary epilepsy in adults, underlying approximately 11% of epilepsy diagnoses.1 The incidence and associations of poststroke epilepsy (PSE) are not well-established; many previous studies have investigated seizure episodes, rather than epilepsy diagnoses, with uncertainty frequently arising from varying definitions of seizure timing, selective recruitment, and short follow-up.2–5
Hemorrhagic strokes, cortical lesions, and large lesions are most consistently associated with poststroke seizures,6 defined as early or late, based on timing and differing pathophysiology.7–10 A wide range (2% to 67%) of patients has been reported for the risk of early or late seizures after ischemic stroke,11 whereas intracerebral and subarachnoid hemorrhages have been associated with a range of between 8% and 15% over varying durations of follow-up.6
There is insufficient evidence to reliably predict those who will have development of PSE and would benefit from prophylactic antiepileptic treatment. Of those with a single seizure, approximately half were observed to have progression to epilepsy in a well-designed study with a maximum follow-up of 6.5 years.3 Associations have been suggested with deep infarctions extending to subcortical structures and late onset of first seizure after ischemic stroke.2,12 PSE impairs self-reported vitality and physical and social functioning.13
We used a population-based register in a multiethnic area of London, including hospital and community diagnoses of first stroke, to investigate the epidemiology and associations of PSE over 12 years, focusing on long-term rates and predictors of epilepsy.
Stroke subjects were recruited from the South London Stroke Register, an ongoing longitudinal register of first-ever strokes in subjects of all ages in a multiethnic inner-city population of 271 817.14
The register’s compilation has been described in detail elsewhere.15 Sixteen overlapping referral sources were used to increase ascertainment. Hospital surveillance for stroke admissions included 2 teaching hospitals within and 3 outside the study area. Patients admitted were identified by daily reviews of acute wards serving stroke patients, weekly checks of brain imaging referrals, and outpatient neurovascular clinic attendance, along with monthly reviews of bereavement records and of bed movement records. All general practitioners within and on the borders of the study area were regularly contacted and asked for notification of all stroke cases. Capture–recapture methodologies suggest that the register’s stroke case ascertainment was an underestimate of approximately 12%.16
Data were prospectively collected by field workers and stroke diagnoses were confirmed by a study physician. Initial data collection was performed as soon as possible after stroke, with formal follow-up at 3 months and annually after diagnosis of first stroke. Follow-up was face-to-face with a field worker or by self-administered questionnaire.
The World Health Organization criteria were used to define stroke.17 Demographic information included age, sex, ethnicity (white, black, other, or unknown), and socioeconomic status based on employment (nonmanual, manual, or unknown). Pathological subtype of stroke was classified as cerebral infarction, primary intracerebral hemorrhage, subarachnoid hemorrhage, or unknown. Infarcts were further defined according to Oxford Community Stroke Project classification (Bamford) definitions as total anterior circulation infarct (TACI), partial anterior circulation infarct, lacunar infarct, posterior circulation infarct, or infarct unspecified.18
Clinical features, including the presence of dysphasia, visual neglect, or visual field defect, were assessed at presentation in addition to case severity indicators comprising Glasgow Coma Scale (GCS) score, the presence of incontinence or dysarthria, and Barthel Index of functional status. Neuroimaging was used to define stroke subtype.
PSE was defined as at least 2 unprovoked epileptic seizures that occurred after the acute phase of the stroke and in the absence of other obvious causes or a history of prestroke epilepsy. The diagnosis of PSE was typically made by a neurologist or epileptologist who, as part of standard clinical practice in the study area, received referrals of patients with new-onset seizures either from emergency departments or from general practitioners.
The first contact with study field workers established the presence of newly diagnosed PSE and censored patients with history of epilepsy before the stroke. Accurate information about how many patients had experienced acute symptomatic seizures, that is, within the first week after the stroke,19 was not available. Therefore, although the patients with PSE diagnosed may include patients who also had acute symptomatic seizures, the proportion of patients with acute symptomatic seizures who had development of PSE could not be ascertained. Subsequent self-administered follow-up questionnaires included the question: Have you been diagnosed by a doctor with epilepsy since last follow-up? A list of medications, including any antiepileptic drugs, was requested at each contact.
Survival time was from date of stroke until death, confirmed by the Office of National Statistics. Sample characteristics were summarized using frequencies and proportions, and characteristics of those who did and did not have development of epilepsy were compared using the χ2 test. Time to first diagnosis of epilepsy was calculated as the time between stroke onset and first follow-up when a diagnosis of epilepsy was reported. Because diagnosis of epilepsy was only recorded at follow-up, exact time was unknown and recorded as the midpoint of the interval between current and previous follow-up. Patients were censored in the interval in which death occurred or, if lost to follow-up, at the last follow-up when no diagnosis was reported.
Kaplan–Meier methods were used to estimate the proportion of patients who had development of epilepsy by 3 months and 1, 5, and 10 years after stroke, and cumulative rates were compared across patient groups using the log rank test. Multivariate complementary log–log models were used to identify all baseline factors that were significantly associated with development of epilepsy poststroke using a forward selection procedure. Analyses were restricted to patients without missing values.
In the analysis it was assumed that patients who missed a follow-up but had no record of epilepsy and were not using an antiepileptic drug at any future or previous follow-ups did not have epilepsy diagnosed by the missed follow-up. A sensitivity analysis was conducted in which all analyses were repeated using data from patients with complete information until development of epilepsy, death, or the 5-year follow-up point to determine robustness of results to these assumptions. All tests were 2 sided and P<0.05 was considered statistically significant. Analyses were conducted using Stata 12MP.
Patients or relatives gave written informed consent to be involved in the study, which was approved by the ethics committees of Guy’s and St Thomas’ NHS Foundation Trust, King’s College Hospital, Queen Square and Westminster Hospital (London).
Three thousand three-hundred seventy-three patients were identified between January 1, 1995 and December 31, 2006, as having first stroke. Sixty-three reported preexisting epilepsy and were excluded. Three thousand three-hundred ten patients, comprising 1673 men and 1637 women, were followed-up for a mean of 3.8 and median of 2.0 years. Two-hundred thirteen subjects (6.4%) had development of PSE. The study population is described fully in Table 1, along with the characteristics of those who did and did not have development of epilepsy; 61.5% of patients with self-reported epilepsy reported use of an antiepileptic drug.
The 10-year cumulative PSE risk was estimated with Kaplan–Meier methods at 12.4% (95% confidence interval [CI], 10.8–14.3). Cumulative risk estimates at 3 months and 1, 5, and 10 years were 1.5% (95% CI, 1.1–2.0), 3.5% (95% CI, 2.9–4.3), 9.0% (95% CI, 7.8–10.5), and 12.4% (95% CI, 10.8–14.3), respectively; further stratification of estimates is presented in Online Table SI. Young age was associated with PSE, affecting 15.6% (95% CI, 12.9–18.7) of patients aged <65 years vs 4.4% (95% CI, 2.0–9.8; P<0.001) aged >85 years; this is shown in comparison with estimates for other age groups in Figure 1. Incidence was associated with stroke subtype, which was highest in 10-year estimates in TACI (28.7%; 95% CI, 21.4–37.7), followed by subarachnoid hemorrhage (21.7%; 95% CI, 13.8–33.2), primary intracerebral hemorrhage (18.2%; 95% CI, 12.9–25.2), partial anterior circulation infarct (13.4%; 95% CI, 10.1–17.9), lacunar infarct (6.4%; 95% CI, 4.5–9.0), and posterior circulation infarct (4.8%; 95% CI, 2.5–8.9; P<0.001), as shown in Figure 2.
All indicators of cortical location were associated with PSE in 10-year estimates, with highest incidence in patients with dysphasia (18.7%; 95% CI, 14.9–23.4) vs without (10.3%; 95% CI, 8.5–12.4; P<0.001), with visual neglect present (24.5%; 95% CI, 19.1–31.0) vs without (9.5%; 95% CI, 7.8–11.5; P<0.001), or visual field defect present (22.0%; 95% CI, 17.3–27.7) vs absent (9.6%; 95% CI, 7.9–11.7; P<0.001).
All case severity indices except dysarthria were associations in univariate analysis of 10-year estimates; these included GCS, with PSE incidence for GCS <8 of 27.8% (95% CI, 18.4–40.6), GCS 9 to 12 of 24.5% (95% CI, 16.4–35.6), and GCS 13 to 15 of 10.8% (95% CI, 9.1–12.7; P<0.001). Incidence in those with incontinence was 20.5% (95% CI, 16.4–25.6) vs 9.3% (95% CI, 7.6–11.4; P<0.001) without, and 15.5% (95% CI, 12.6–18.9) for those with moderate–severe scores on the Barthel Index, 12.6% (95% CI, 8.7–18.1) for those with mild disability, and 8.2% (95% CI, 5.9–11.4; P<0.001) for independent patients. Dysarthria, sex, ethnicity or socioeconomic status were not significantly associated with PSE in 10-year estimates.
Multivariate analysis of independent associations of PSE is presented fully in Table 2. Advanced age was independently associated with a reduced risk of PSE; in subjects aged >85 years, 10-year hazard ratio (HR) was 0.21 (95% CI, 0.06–0.66) compared with patients aged <65 years (reference group; P=0.004).
Stroke subtype was independently associated with PSE. Compared with TACI (reference group), epilepsy risk was significantly lower in lacunar infarct (HR, 0.46; 95% CI, 0.23–0.89) and posterior circulation infarct (HR, 0.27; 95% CI, 0.11–0.67; P=0.020).
Dysphasia (HR, 1.48; 95% CI, 1.01–2.17; P=0.044), visual neglect (HR, 1.81; 95% CI, 1.20–2.73; P=0.006), or incontinence (HR, 1.53; 95% CI, 1.05–2.21; P=0.027) were independently associated with PSE. The presence of dysarthria was independently associated with a reduced risk of PSE (HR, 0.68; 95% CI, 0.48–0.97; P=0.032).
Separate analyses censoring patients at 5-year follow-up or including only patients with complete data up to 5 years after stroke showed no differences in findings of univariate or multivariate analyses when compared with results obtained using all available data.
This is the first population-based study to investigate the incidence and associations of PSE with a follow-up of 12 years and large sample size. We have clearly established that the risk for PSE continues to increase considerably outside the acute phase, a finding of importance given the high variability in results of previous studies.20
We report a significant independent association of stroke subtype with PSE, with incidence highest in 10-year estimates in TACI followed by subarachnoid hemorrhage, primary intracerebral hemorrhage, partial anterior circulation infarct, lacunar infarct, and posterior circulation infarct. This broadly supports and adds to previous studies investigating poststroke seizures only. Furthermore, our findings support the previously observed relationship of cortical insults (such as TACI) and subsequent epilepsy.6
The temporal lobe is the most epileptogenic region of the brain and the most frequent subject of epilepsy surgery,21 followed by the frontal lobe, both with anterior circulation supply. Large events involving the middle and superior temporal gyri are particularly associated with the development of late seizures,22 and cortical tumors of any type are most associated with epilepsy.23 All 3 markers suggestive of cortical stroke were significantly associated with PSE on univariate analysis, whereas multivariate analysis showed significant independent associations of PSE with visual neglect and dysphasia.
The association of young age with PSE is less well-established by previous studies, which were limited by shorter follow-up and investigation of poststroke seizures rather than epilepsy diagnoses. Young age was an associated factor in a retrospective study of 6044 patients with first or recurrent stroke and in a large, hospital-based, retrospective cohort study of 1880 patients with first stroke.4,24 Conversely, the association of seizures with old age has been reported, for instance, in a retrospective population-based study of 675 British patients with a minimum follow-up of 2 years.3 In relation to our findings, we propose that younger patients are more likely to experience large cortical infarcts rather than small vessel disease–related lacunar events, and that this may contribute to subsequent epilepsy development. The data may be confounded by differing rates of presentation; patients with less cognitive impairment may be more likely to report seizures and to reach diagnosis.
Incontinence at presentation was significantly associated with PSE, independent of other variables investigated. This is used along with dysarthria, GCS, and Barthel Index of functional status in the South London Stroke Register as a measure of stroke severity. A worse GCS score and Barthel Index were each associated with PSE on 10-year estimates. These findings are suggestive of an association of PSE with more severe stroke.
In our multivariate analysis, the absence of dysarthria was associated with a significant increase in risk of PSE, in apparent contradiction to previous studies. This may be explained by the multifactorial pathogenesis of dysarthria, which is known to arise from cerebellar, cortical, or subcortical insults. Only 6% of patients with dysarthria have lesions in the lower part of the primary motor cortex; indeed, subcortical and brain stem lacunar lesions are more associated with the development of dysarthria than cortical events.25 Because TACIs comprise only 17% of all ischemic strokes,18 it might be expected that dysarthria, occurring more frequently in the remaining subtypes, would be negatively associated with PSE.
The strengths of this study were its population-based design that followed-up patients presenting with first stroke in the community and in hospital. The area of South London from which recruitment took place is ethnically diverse and no differences were found between ethnic groups. Findings are therefore unlikely to be specific to one or another group and are generalizable. We included patients from the beginning of the study and those subsequently recruited, with a mean follow-up period approaching 4 years, which is considerably longer than many previous studies of poststroke seizures. The study is of value because of its focus on epilepsy diagnosis rather than poststroke seizures, which are already extensively characterized, with highly variable findings leading to difficulty in interpretation for patients and clinicians.
There were several weaknesses in the design of the study. Epilepsy diagnoses at follow-up contacts were reported by patients and not confirmed by study physicians. Therefore, some overestimation of the incidence of PSE cannot be ruled out because other causes of late-onset epilepsy may have escaped detection at follow-up assessments. Minor underestimation is also possible because some patients with mild sensory ictal symptoms only, such as epigastric sensation or déjà vu, may not have sought medical advice. It is also possible that excluding patients with preexisting epilepsy, who may have development of acute symptomatic seizures or PSE, resulted in a slight underestimation of the true incidence of PSE. The net effect of these types of error would, however, seem small and unlikely to significantly affect our results given the substantial study population size. Finally, the study design did not allow estimation of the incidence of acute symptomatic seizures and the associated risk of subsequent development of PSE.
Not all patients at each follow-up stage completed the self-assessment questionnaire. Time of diagnosis was defined as the period before the first follow-up at which a diagnosis of epilepsy was reported, with patients who missed follow-ups censored at the last follow-up at which antiepileptic drugs were not prescribed if they did not report a new diagnosis of epilepsy. This could introduce bias into the findings producing an underestimate of prevalence if a diagnosis was made during a period in which the patient did not complete any follow-ups. However, sensitivity analysis using patients with complete data up to 5 years after stroke showed no differences in findings on univariate or multivariate analyses.
This population-based study shows PSE to be independently associated with young age and large, hemorrhagic, and cortical lesions. The findings are of use in the risk stratification of patients with acute stroke and may help to inform patients and clinicians about the risk of future epilepsy and to aid the design of future trials of prophylactic antiepileptic medication in patients with stroke.
The authors thank all of the patients and their families involved, in addition to the health care workers, field workers, and supporters who have contributed to the South London Stroke Register since 1995.
Sources of Funding
This project received funding from the Stroke Association and National Institute for Health Research (NIHR) Program Grant funding (RP-PG-0407-10184). The South London Stroke Register also has received funding from the Northern and Yorkshire National Health Service (NHS) R&D Program in Cardiovascular Disease and Stroke, the Guy’s and St Thomas’ Hospitals Charitable Foundation, the Stanley Thomas Johnson Foundation, the Stroke Association, and Department of Health (DH) HQIP funding. This article presents independent research commissioned by the NIHR under its Program Grants for Applied Research funding scheme (RP-PG-0407-10184). The views expressed in this article are those of the authors and not necessarily those of the NHS, the NIHR, or the DH. The authors acknowledge financial support from the DH via the NIHR comprehensive Biomedical Research Center award to Guy’s & St Thomas’ NHS Foundation Trust in partnership with King’s College London and King’s College Hospital NHS Foundation Trust.
M. Koutroumanidis has received research support, funding for travel, and speaker honoraria from UCB, and funding for travel from Eisai.
- Received November 19, 2012.
- Accepted December 3, 2012.
- © 2013 American Heart Association, Inc.
- Burn J,
- Dennis M,
- Bamford J,
- Sandercock P,
- Wade D,
- Warlow C
- Sun DA,
- Sombati S,
- DeLorenzo RJ
- Camilo O,
- Goldstein LB
- Gupta SR,
- Naheedy MH,
- Elias D,
- Rubino FA
- 14.↵Office for National Statistics. 2001 Census: Standard Area Statistics (England and Wales) [computer file]. ESRC/JISC Census Programme, Census Dissemination Unit, Mimas (University of Manchester). HMSO. 2001.
- Stewart JA,
- Dundas R,
- Howard RS,
- Rudd AG,
- Wolfe CD
- Tilling K,
- Sterne JA,
- Wolfe CD
- Krämer G
- Engel J Jr.,
- Wiebe S,
- French J,
- Sperling M,
- Williamson P,
- Spencer D,
- et al