Poststroke epilepsy: update and future directions

Abstract

Stroke is among the most common causes of epilepsy after middle age. Patients with poststroke epilepsy (PSE) differ in several respects from patients with other forms of structural–metabolic epilepsy; not least in age, age-related sensitivity to side effects of antiepileptic drugs (AEDs), and specific drug–drug interaction issues related to secondary-stroke prophylaxis. Encouragingly, there has lately been remarkable activity in the study of PSE. Three developments in PSE research deserve particular focus. First, large prospective trials have established the incidence and risk factors of PSE in the setting of modern stroke care. Stroke severity, cortical location, young age, and haemorrhage remain the most important risk factors. Second, although more studies are needed, epidemiological data indicate that the risk of PSE may be influenced, for instance, by statin treatment. Third, studies are emerging regarding the treatment and prognosis of PSE. Levetiracetam and lamotrigine may be well tolerated treatment options and seizure freedom is achieved in at least a similar proportion of patients as in other epilepsies. Furthermore, new animal models such as photothrombotic stroke gives hope of a more clear understanding of PSE epileptogenesis in the near future. In summary, PSE shows indications of maturing into an independent epilepsy research field. This review summarizes recent advances in our understanding of PSE and provides an update on management issues such as diagnosis, AED selection, and prognosis. Finally, future research challenges in the field are outlined.

Introduction

Cerebrovascular illness is a major cause of epilepsy after middle age [Forsgren et al. 2005; Syvertsen et al. 2015]. Over the last decade or so, there have been remarkable improvements in the management of stroke, with advances in emergency revascularisation being paralleled by raised ambitions in secondary prophylaxis, rehabilitation, and management of nonmotor sequelae such as fatigue. Poststroke epilepsy (PSE) thereby occurs in an optimistic medical context, but advances in management of poststroke seizures have not quite matched those seen in other stroke treatment domains. In fact, guidance was until recently scarce for neurologists pondering relatively fundamental management issues, such as when to treat whom with what. Fortunately, there seems to be a growing research interest in PSE. This is perhaps part of a general trend in the epilepsy field of increased focus on aetiology. Traditionally, patients with structural–metabolic epilepsy of widely different origins, from developmental tumours to stroke, were studied under the umbrella term of ‘partial epilepsy’. Although pathways of epileptogenesis may converge, it is clear that structural–metabolic epilepsies of different aetiologies occur in different contexts, medically and demographically, and that management considerations differ in a 30-year old with a low-grade glioma from a 60-year old with hemiparesis and atrial fibrillation.

For clinicians and patients, the focus on aetiology is therefore good news. Better stratification of patients in studies according to underlying aetiology is likely to result in more relevant knowledge and improved management. This review aims to summarize some of the recent advances in PSE research and address some important questions clinicians are likely to face, such as risk factors for PSE after stroke, the incidence of PSE in the setting of modern stroke care, and an update on current management of PSE. Not all advances have been based on new scientific data. Also covered are policy documents from the International League Against Epilepsy (ILAE) on the definition of acute symptomatic seizures and the definition of epilepsy [Beghi et al. 2010; Fisher et al. 2014], which have spread expert practice and brought valuable structure to the management of PSE.

Epileptogenesis, early and late seizures

PSE shares features with other forms of structural–metabolic epilepsy. As in, for instance, post-traumatic epilepsy, there is often a latent phase after the insult during which the brain is thought to undergo epileptogenesis and acquire a predisposition for seizures. Among proposed epileptogenic mechanisms are inflammation and remodelling of synaptic networks, perhaps influenced by genetic susceptibility [Silverman et al. 2002; Pitkanen et al. 2015]. Historically, PSE has been difficult to model in rodents, and animal researchers have therefore mainly focused on models of epilepsy after trauma, electric kindling, chemoconvulsant-induced status epilepticus (SE), or cortical injections of irritants such as tetanustoxin or ferrous chloride. Relatively recently, photothrombotic stroke has been established as a model for PSE in rodents [Kelly et al. 2001; Pitkanen et al. 2007], so whether PSE differs from other forms of acquired epilepsy in the most basic mechanisms of epileptogenesis will most likely be elucidated in the near future.

The concepts of latency and epileptogenesis form an important basis for clinical understanding of the concept of early versus late seizures. Early seizures occur immediately after a stroke, and are thought to be consequences of local metabolic disturbances that have not necessarily altered neuronal networks, but have made them epileptic in nature. Late seizures occur when epileptogenesis is postulated to have occurred and the brain has acquired a predisposition for seizures [Silverman et al. 2002]. As expected from this model, early and late seizures carry different risks of seizure recurrence (Figure 1). However, the concept of a latent phase and epileptogenesis reflects our current understanding of epilepsy and may be oversimplified [Loscher et al. 2015]. Some patients with early seizures will develop PSE, and future research, for instance, on biomarkers, may improve our understanding of these cases and aid in their early identification.

Figure 1.

Risk-stratification flow chart and treatment algorithm for seizures (that are not considered acute or symptomatic for other reasons) after ischaemic stroke, intracerebral haemorrhage or subarachnoid haemorrhage.

Most risk estimates are from selected observational studies often influenced by AED treatment, varying time limits for early and late seizures, and varying definitions of epilepsy.

AED, antiepileptic drugs; ICH, intracerebral haemorrhage; IS, ischaemic stroke; SE, status epilepticus; HT, haemorrhagic transformation; PSE, poststroke epilepsy; SAH, subarachnoid haemorrhage; sz, seizure.

Epidemiology

The reported long-term cumulative risk of PSE after a cerebrovascular event varies between 2% and 15% (Table 1). The variations reflect differences in study cohorts regarding stroke aetiology or severity and in study methodology such as follow-up time, outcome measures, whether patients remained in the study after subsequent strokes, definitions of late and thus unprovoked seizures, and whether survival was corrected for. Interpretation of observational studies requiring two unprovoked seizures for diagnosis of epilepsy are also complicated by the fact that many patients most likely started AED treatment after the first late seizure.

Table 1.

Summary of selected studies on incidence of poststroke epilepsy published in the last 20 years.

Study	n	Study type	Percentage IS/ICH/SAH	Follow up (mean or total)	Outcome measure	Cumulative risk
So et al. [1996]	535	Retrospective	100/0/0	5.5 yrs	PSE: 2 sz	IS: 8.9% 10-yr risk
Burn et al. [1997]	675	Retrospective	81/10/5	>2 yrs	All seizures > 24 h	All: 11.5%, IS 9.7%, ICH 26.1%, SAH 34.3% (5-yr actuarial risk)
Paolucci et al. [1997]	306	Prospective (rehab inpatient)	81/19/0	No data	1 LS	All: 15% IS: 12% ICH: 27%
Bladin et al. [2000]	1897	Prospective	86/14/0	9 mths	PSE: 2 sz	All: 2.5%, IS: 2.1%, ICH: 2.6%
Lossius [2002]	482	Prospective (age > 60)	92/8/0	12 mths	PSE: 2 sz after 4 weeks	All: 2.5% PSE
Lamy et al. [2003]	581	Prospective (age 18–55)	100/0/0	37.8 mths	PSE: 2 sz	IS: 2.3% 3-yr risk
Kammersgaard and Olsen [2005]	1195	Prospective	93.7/6.3/0	7 yrs	PSE: 2 sz and need of AED	All: 3.2%
Okuda et al. [2012]	448	Retrospective (rehab inpatient)	64/36/0	8 mths	PSE: 2 sz	All: 1.3%
Arntz et al. [2013b]	697	Prospective (age 18–50)	93 incl TIA/7/0	9.1yrs	PSE: 2 sz	All: 7%, IS: 8%, TIA: 4%, ICH 23%
Graham et al. [2013]	3310	Prospective	73.4/13.8/5.7	3.8 yrs	PSE: 2 sz	All: 12.4%, IS: 4.4–28.7%, ICH: 18.2%, SAH: 21.7% estimated 10-yr risk
Jungehulsing et al. [2013]	1020	Prospective	87/8.7/3.8*	2 yrs	LS > 1 week	All: 8.2% 2-yr risk
Qian et al. [2014]	935	Prospective	0/100/0	1–16 yrs	PSE: dx in records	7%
Guo et al. [2015]	1832	Prospective	100/0/0	2.5 yrs	PSE: 2 sz	IS: 5%
Bryndziar et al. [2015]	489	Retrospective	100/0/0	6.5 yrs	LS > 2 week	IS: 4.4% LS
Serafini et al. [2015]	782	Prospective	79.28/14.83/3.2	2 yrs	PSE: 2 sz	All: 2.22%, IS 1.97%, ICH: 4.35%
Huttunen et al. [2015]	876	Retrospective	0/0/100	76 mths	PSE: 1 LS > 1 week	SAH: 12% 5-year risk of LS and AED

^*Aetiology data only provided for 725 survivors.

IS, ischaemic stroke; ICH, intracerebral haemorrhage; SAH, subarachnoid haemorrhage; PSE, poststroke epilepsy; AED, antiepileptic drug; LS, late seizures; sz, seizures; dx, diagnosis.

Both early and late seizures are more common after haemorrhage than after infarctions, with the exception of total anterior circulation infarctions, which seem to carry an even higher risk than haemorrhages of PSE [Burn et al. 1997; Bladin et al. 2000; Labovitz et al. 2001; Procaccianti et al. 2012; Pezzini et al. 2013; Serafini et al. 2015]. An early study reported an incidence of late seizures in 1000 patients with nonembolic cerebral infarction of 2.7%, after an average follow-up time of 6 to 12 months [Louis and McDowell, 1967]. This is similar to later data [Burn et al. 1997; Bladin et al. 2000; Kammersgaard and Olsen, 2005]. Interestingly, different investigators recently described quite high long-term cumulative risks. In the largest study to date that was performed in the UK and followed 3310 patients with newly diagnosed stroke for an average of 3.8 years, the 10-year risk of PSE was calculated to be 12.4% [Graham et al. 2013]. In a smaller, but population-wide study, 4.4% of 481 patients developed late seizures over an average follow-up period of 6.3 years [Bryndziar et al. 2015]. Another study on 1020 patients reported a 2-year risk of 8.2% [Jungehulsing et al. 2013]. Finally, 5.0% of 1832 prospectively followed patients with ischaemic stroke were reported by Guo and coworkers to have developed PSE after 2.5 years [Guo et al. 2015]. Subarachnoid haemorrhage is less common than other stroke subtypes, and so accounts for a small proportion of all cases of PSE, but seems to carry a substantial risk of PSE [Kotila and Waltimo, 1992; Burn et al. 1997; Graham et al. 2013; Huttunen et al. 2015].

Predictors

In almost all larger studies on the risk of PSE, multivariate analyses were performed to identify independent risk factors (Table 2). Such predictors include stroke severity, cortical symptoms, haemorrhage, total anterior circulation infarcts, young age at stroke, and early seizures [Bladin et al. 2000; Graham et al. 2013; Jungehulsing et al. 2013]. The strengths of associations between PSE and many of the variables differ across studies, probably for methodological reasons. The findings regarding early seizures are especially difficult to assess. Early seizures were, for instance, associated with an increased risk of PSE in the Copenhagen stroke study [Kammersgaard and Olsen, 2005], but since they were defined as seizures occurring within 2 weeks, some of the early seizures might in fact have been late seizures according to the current definition [Beghi et al. 2010]. Nonetheless, the findings are of interest in light of the current discussion on whether the latent phase is perhaps not latent at all, but in some circumstances, harbours early epileptogenesis [Loscher et al. 2015]. With the possible exception of certain evacuated intracranial haemorrhage (ICH) [Qian et al. 2014], none of the clinical risk factors identified to date indicates a seizure risk high enough to warrant prophylactic antiepileptic treatment, although practice may vary.

Table 2.

Independent predictors of poststroke epilepsy in selected studies.

Study	Predictors in all patients (mixed aetiology)	Predictors in IS	Predictors in ICH	Comment
Paolucci et al. [1997]	Putaminal haemorrhage Lobar haemorrhage Severity of stroke Young age (<46 yrs)			Rehab inpatients
Graham et al. [2013]	Young age (<65 yrs) Stroke subtype Cortical signs Stroke severity
Bladin et al. [2000]	Late seizures (>2 weeks)	Late seizures
Kammersgaard and Olsen [2005]	Young age Onset stroke severity Lesion size Haemorrhagic stroke Early seizure
Serafini et al. [2015]	Young age Cortical location
Lossius [2002]	Stroke severity
Lamy et al. [2003]		Early seizure Cortical sign Lesion size		In young adults with cryptogenic stroke
Qian [2014]			Subcortical location Age (inversely) Hematoma evacuation	Predictors of late seizures

IS, ischaemic stroke; ICH, intracerebral haemorrhage.

Some predictors have been indirectly examined by studies on high-risk populations. Regarding young age, Arntz and coworkers prospectively followed 697 transischaemic attack (TIA) or ischaemic stroke patients aged 18 to 50, and calculated the cumulative risk of epilepsy to be 7% [Arntz et al. 2013b]. Lower rates were found by Lamy and coworkers [Lamy et al. 2003], but the authors state that the study design may have excluded some high-risk patients. Nonetheless, it is entirely possible that the increased risk attributed to young age reflects survival rather than increased susceptibility to epileptogenesis [Hauser, 2013]. Regarding stroke severity, the incidence of seizures among 306 patients admitted to a rehabilitation facility because of stroke sequelae was 15%, with recurrence in 90% within a year, in spite of AED treatment [Paolucci et al. 1997]. Another study reported lower risks, but had short follow-up time and possible selection bias for patients with less severe strokes [Okuda et al. 2012].

Prevention

Since stroke is often readily recognizable and since clinical risk factors allow selection of high-risk patients, PSE is an interesting area for studies on preventing epilepsy. Performing antiepileptogenesis clinical trials has, however, proven difficult. In general, since AEDs have side effects and work by modulating neuronal signalling, the risk of such drugs hampering rehabilitation after stroke is not negligible and poses an ethical dilemma. Furthermore, stroke patients are likely to have additional strokes, which might be epileptogenic and make it difficult to establish if a treatment reduces the risk of stroke or epilepsy. Nonetheless, some attempts have been made to prevent PSE in high-risk populations. In patients with ICH, valproic acid treatment for 1 month reduced the number of early seizures, but had no impact on the development of epilepsy [Gilad et al. 2011]. Diuretics, like thiazides and furosemide, protect against seizures in animal models, and in observational epidemiological studies seem to reduce the risk of seizures in humans [Hesdorffer et al. 1996, 2001]. Other candidate drugs with potentially antiepileptogenic properties include levetiracetam (LEV) and statins [Temkin, 2001; Pitkanen et al. 2007]. A nested case-control study in Canada on patients with cardiovascular illness showed that statin use reduced the risk of subsequent hospitalization for epilepsy [Etminan et al. 2010]. The study elegantly used other drugs without known antiepileptogenic properties as controls, and found no benefits associated with their use. Similar findings were recently reported from a prospective observational study, where statin use was associated with a lower risk of early seizures. Among patients with early seizures (indicating a higher than average risk of PSE), statin use was associated with reduced risk of epilepsy [Guo et al. 2015]. The authors did not provide subgroup analyses for seizure types and the observational design of the study is an important limitation [Siniscalchi, 2015]. The data on statins are intriguing and more studies are needed. In mice, different statins seem to have different anticonvulsant potential (perhaps related to variations in lipophilicity) and the doses required for anticonvulsant effects suggest that the antiepileptogenic effect may not necessarily be linked to HMG-CoA reductase inhibition [Russo, 2013]. For all candidate compounds, more investigations regarding antiepileptic properties are required, which may prove challenging. In perhaps the most interesting paper in the reference list, Van Tuijl and coworkers describe the difficulties encountered in a trial of LEV as antiepileptogenic therapy after stroke [van Tuijl et al. 2011].

Intuitively, the most direct way to prevent PSE would be to prevent or mitigate the initial stroke. To date, there is, however, little evidence that improved stroke care reduces the incidence of PSE. In a study on 257 patients treated with intravenous thrombolysis, the rate of late epileptic seizures was 11.3% [Gensicke et al. 2013]. The figure is very similar to that seen in another study that retrospectively compared cases that had received t-PA with controls who had not, and found no significant difference in epilepsy incidence at 2 years (10.8% for t-PA versus 8.0% for controls). This study was perhaps underpowered given the fact that cases and controls differed in several risk factors for PSE, such as stroke severity and haemorrhagic transformation [Tan et al. 2012]. Other interventions in the acute setting that might mitigate brain damage, such as blood-pressure management, different antithrombotic treatments, and acute high-dose statin treatment have not been assessed for effect on PSE risk.

Diagnosis and treatment decision

According to the new ILAE practical definition, epilepsy can be diagnosed after a single seizure if other findings support that there is a risk of seizure recurrence equivalent to that seen after two unprovoked seizures [Fisher et al. 2014]. Since patients with a late-unprovoked seizure after stroke have a 71.5% (95% CI: 59.7–81.9%) risk of another unprovoked seizure, they fall into this category [Hesdorffer et al. 2009]. The new definition is not uncontroversial, mostly due to conflicting data on the actual recurrence risk of seizures after a first late-poststroke seizure and whether a diagnosis benefits patients [Hauser, 2013].

There is no evidence to date that treatment with AEDs prevents the development of PSE and the risk of seizure after stroke is relatively low, so primary prevention is not deemed appropriate [Labovitz et al. 2001; Temkin, 2001; Gilad et al. 2011; Sykes et al. 2014; Serafini et al. 2015], except perhaps for some cases of lobar ICH [Steiner et al. 2006; Qian et al. 2014]. A single early seizure does not typically warrant seizure prophylaxis in ischaemic stroke, but short-term treatment is often initiated if multiple early seizures occur or after a single seizure in cases of ICH or haemorrhagic transformation [Beleza, 2012; Qian et al. 2014]. Practice seems to vary. Early seizures may be a risk factor for late seizures, at least in ICH, but the risk is not described as higher than that seen after a single unprovoked seizure [Kilpatrick et al. 1992; So et al. 1996; Qian et al. 2014; Serafini et al. 2015], making AED withdrawal reasonable in most cases. In select cases, prolonged prophylaxis may be warranted. For instance, the 10-year cumulative risk of subsequent unprovoked seizures seems to increase two- to threefold in patients with SE compared with patients with self-terminating first seizures, but it is not clear if patients with stroke were significant contributors to the increased risk in those studies [Hesdorffer et al. 1998, 2009]. Worryingly, one study demonstrated a very high recurrence risk after SE after stroke and substantial severity of the recurrent seizures [Rumbach et al. 2000]. No significant difference could be detected in recurrence risk between early and late SE, perhaps due to the low number of patients. More data is needed, but meanwhile a pragmatic approach may be to offer seizure prophylaxis for a longer period of time, although perhaps not indefinitely, since most patients with PSE will have seizures recurrence within 1–2 years, thereby confirming the diagnosis [Bryndziar et al. 2015; Guo et al. 2015]. Generally, patients should always to be involved in the discussion on stopping AEDs. Figure 1 includes some recent risk estimates for different patient categories and may aid in risk stratification and counselling of patients after early seizures.

In the case of a late seizure, the patient should be informed of a high recurrence risk and offered seizure prophylaxis. Individual factors such as constant supervision, lack of ambulation and therefore low risk of seizure-related injury, very mild seizures, etc. may be reasons to defer treatment. Factors that may help in determining if a seizure is indeed postapoplectic are semiotic in relation to the lesion and latency after the stroke, typically less than 2–3 years [Bryndziar et al. 2015; Guo et al. 2015].

There are currently very little specific data to guide management regarding seizures after the less common stroke type central sinus venous thrombosis (CSVT) [Price et al. 2014]. Pending more specific data, management follows that of other forms of PSE, but the distinction between early and late seizures can be more difficult in CSVT, and clinical judgement is often the only available resource.

AED treatment

When a treatment decision has been made, drug selection is made in the same manner as for other forms of epilepsy; tailored to each patient based on assumed efficacy, concurrent medication, and side-effect profile. Traditional choices for PSE have been CBZ or PHT [Silverman et al. 2002], but as discussed below other AEDs are today probably more suitable choices for many patients. Regarding partial-onset seizures in adults in general, the latest ILAE report on efficacy of AEDs states that level A evidence for efficacy exists for carbamazepine (CBZ), LEV, phenytoin (PHT), and zonisamide (ZNS). For elderly patients with partial-onset seizures, gabapentin (GBP) and lamotrigine (LTG) have level A evidence [Glauser et al. 2013].

Some low-level evidence has accumulated on treatment of PSE specifically. With customary restraint, a Chochrane review mentions LTG and LEV as interesting treatment options, but notes lack of sufficient data to allow generalizations [Sykes et al. 2014]. In the SANAD trial on partial epilepsy [Marson et al. 2007], GBP seemed slightly less effective than other drugs, but may perhaps have a role in PSE. In a small uncontrolled trial on 71 elderly patients with a first late seizure, GBP was well tolerated in all patients and seizures recurred in only 18.3% during the mean follow up of 30 months [Alvarez-Sabin et al. 2002]. A more modest rate of seizure freedom was noted by Gilad and coworkers in a small but randomized trial with 64 patients with late poststroke seizures comparing CBZ and LTG [Gilad et al. 2007]. Seizure freedom was achieved in 44% and 72%, respectively. The difference in efficacy was not statistically significant, but the difference in withdrawal due to side effects was, with LTG being better tolerated [Gilad et al. 2007]. Another randomized open-label trial on 128 patients compared LEV and slow-release CBZ for PSE. The study demonstrated no significant difference in efficacy, but cognitive side effects were less frequent in the LEV group, highlighting a potential advantage of this drug over CBZ [Consoli et al. 2012]. However, effect measurements were small (seizure freedom was achieved in 94% of patients treated with LEV and 85% of patients treated with CBZ) and the authors note that the study was initially designed for a much larger population. Recently, LEV was once again shown to have better tolerability than controlled-release CBZ in an elderly population with epilepsy of predominantly cerebrovascular aetiology [Werhahn et al. 2015]. LTG performed closely to LEV in the same study, but was not statistically significantly different from the LEV or CBZ.

With all AEDs, and especially enzyme inducers, one should consider the potential for drug interactions with secondary-stroke prophylaxis and the effects of the AED itself on the vascular risk profile of the patient. The combination of strong enzyme inducers, exemplified by CBZ and PHT, with new anticoagulants like apixaban or dabigatran is discouraged in the prescription information. CBZ lowers the exposure to concomitant simvastatin, as do newer generation AEDs with enzyme-inducing properties like ESL [Falcao et al. 2013; Vyas et al. 2015]. Regarding vascular risk, AEDs can, for instance, impact serum lipids, weight and risk of cardiac arrhythmias [Katsiki et al. 2014]. The literature on the effect of AEDs per se on lipids is not unequivocal. Negative effects seems larger for older AEDs, but this may reflect publication bias or lack of data, as noted in a recent systematic review whose authors recommend vigilance regarding vascular risk-factor modification in patients on AEDs [Vyas et al. 2015]. As a rule, treatment with AEDs in elderly patients combines the described challenges of drug–drug interactions and cardiovascular risk with age-related changes in pharmacokinetics and increased susceptibility to cognitive side effects in particular [Siniscalchi, 2012]. Regarding cognitive function, LTG, LEV, and GBP seem to have fewer adverse cognitive effects than CBZ [Park, 2008].

Based on the above studies, this author considers LEV, LTG, and sometimes GBP, as reasonable first choices in PSE. In cases of severe or frequent seizures, GBP would probably not be a first choice given the SANAD efficacy results [Marson et al. 2007]. There are insufficient data to discourage the use of CBZ or other sodium-channel blockers, but drug interactions as well as enzyme-inducing and pro-arrhythmogenic properties of at least older compounds make them less then ideal in older PSE patients. The role of new AEDs such as oxcarbazepine, eslicarbazepine, and lacosamide remains to be determined. The side-effect profiles of many of these drugs seem encouraging, also for the newer sodium-channel blockers, which may be less enzyme inducing than their older relatives. In all cases of AED treatment, country-specific regulations and licensing need to be observed and may limit physician choice.

Epilepsy prognosis

PSE is often described as an easily manageable form of epilepsy, where monotherapy suffices to control seizures [Silverman et al. 2002; Bryndziar et al. 2015]. If so, recent real-life data indicate that current management of PSE may not be ambitious enough. In a recent small study, we retrospectively described the clinical course of poststroke epilepsy in 36 patients. Approximately two thirds achieved sustained seizure freedom (crudely measured as seizure free at last follow up), which is similar to the figures seen in patients with epilepsy in general [Zelano et al. 2015]. Other studies on small populations report varying results [Paolucci et al. 1997; Bryndziar et al. 2015]. Taken together, there may be room for improvement in management that should aim for seizure freedom in the same manner as in other forms of epilepsy. More prospective data are needed on properly sized cohorts for proper estimates of achieved seizure freedom in general and for PSE after different stroke subtypes.

Outcome

Compared with the information on the risk of developing PSE, there is less information on risks associated with the condition once it is established. Regarding mortality, PSE seems to have an impact at least in young patients. In patients aged 18–50, epilepsy was associated with increased mortality also after adjustment for confounders [Arntz et al. 2015]. Short-term mortality at 30 days had an HR (hazard ratio) of 4.8 (95% CI: 1.7–14.0) and long-term mortality had an HR of 1.8 (95% CI: 1.2–2.9). This increased risk may not extend to PSE patients in general. A large recent study failed to identify PSE as an independent risk factor of 2-year mortality, and similar results have been reported by other investigators [Dhanuka et al. 2001; Serafini et al. 2015]. Conflicting results exist [Hesdorffer et al. 2009]. Future larger studies are needed to assess the actual impact of PSE on mortality in different patient groups, with stroke severity and comorbidities taken into account. Intuitively, one would perhaps expect the impact of seizures per se on mortality to be greater in younger patients, where nonseizure-related mortality is otherwise not very high. For elderly, the influence of AED treatment on overall vascular risk may be of greater importance than seizure-related risks.

Regarding rehabilitation, function in 537 young patients with PSE was evaluated by Arntz and coworkers. In a multivariate analysis, PSE was independently associated with poor outcome after cerebral infarction, as assessed by mRs (modified Rankin scale) [Arntz et al. 2013a]. This is perhaps not generalizable to all PSE patients, as another study on rehabilitation patients did not find a significant association between seizures and rehabilitation parameters [Paolucci et al. 1997]. A large study demonstrated that early seizures do not influence functional outcome at 6 months after ICH [De Herdt et al. 2011], but as is often the case in observational studies, interpretation is difficult since AED treatment was started in a proportion of patients based on clinician judgement. Little data exist on the impact of PSE on quality of life.

Prestroke epilepsy

Since stroke is a common cause of seizures after middle age, an interesting question is whether new onset seizures in this age group warrant stroke work up and prevention. The occurrence of seizures prior to a cerebrovascular event was noted in the Oxfordshire community stroke project [Burn et al. 1997] and a more recent UK database study on 4709 individuals and an equal number of matched controls found that the HR for subsequent stroke in people experiencing onset of seizures after the age of 60 was 2.89 (2.45–3.41)[Cleary et al. 2004]. Similarly, Chang and coworkers investigated the risk of stroke in people with epilepsy in Taiwan and found a threefold increased risk of stroke during the study period. Presumably, disease mechanisms vary between age groups [Chang et al. 2014]. In younger age groups, genetic vulnerability, accelerated atherosclerosis due to AED treatment, or cardiovascular changes due to seizures might contribute to a higher risk of stroke, whereas in older patients, seizures may be a marker of cortical lesions due to already existing but subclinical cerebrovascular disease. More research is needed on the importance of stroke prophylaxis in elderly patients with newly diagnosed epilepsy. Such information is especially needed given the potentially negative impact on vascular risk exerted by AEDs [Katsiki et al. 2014; Vyas et al. 2015].

Future directions

With increased specialization in healthcare, neurologists will most likely face increased demands to participate in the management of PSE patients, a group traditionally seen by generalist physicians. There will certainly be structural challenges in providing access to neurological care for PSE patients and ideally, team-based solutions should be developed to tackle the complexity of comorbidities and increasingly complex secondary-stroke prophylaxis regimes.

Regarding prevention, biomarkers of epileptogenesis are most likely of vital importance. Studies on prevention of PSE in general stroke populations are difficult to perform, even with ambitious large multicentre efforts [Van Tuijl et al. 2011]. Inclusion of patients in studies would probably be facilitated if high-risk individuals could be identified, a task for which current clinical predictors are too nonspecific. Biomarkers enabling identification of epileptogenesis and thereby selection of at-risk patients therefore seems to be a vital first step for development of future preventive treatments [Pitkanen et al. 2015]. Advances in neurochemistry, neuroimaging and neurophysiology have greatly improved our abilities to detect brain pathology and a number of modalities should be evaluated for potential as biomarkers. Biomarkers for epileptogenesis may include blood tests for brain injury or inflammation, MRI (magnetic resonance imaging) or functional MRI, high-frequency oscillations, transcranial magnetic stimulation, magnetic encephalography, or combinations of the above. Large-scale application of these methods in patients may be too expensive and laborious for immediate studies, and should probably be preceded by a deeper understanding of epileptogenesis based on refined animal models.

In the near future, prevention of PSE is perhaps not the most pressing issue. For patients of today, prospective multicentre studies are needed to clarify both prognosis and aspects of management of PSE. Future studies should aim to clarify the impact of AED treatment on vascular-risk profile and rehabilitation. Another question is whether poststroke seizures can cause brain damage and thereby deterioration in neurological function, as is suggested by diffusion weighted imaging changes described in PSE patients with long-lasting neurological deterioration after seizures [Kumral et al. 2013]. The impact of repeat early seizures on epileptogenesis after stroke is also not clear. Data are not only needed for improvement of management, but also for proper allocation of healthcare resources.

In summary, the last years have seen remarkable research activity on PSE. Hopefully, the impressive activity in the field will continue and lead to substantial improvement in the care of an important patient group.