John Libbey Eurotext

Epileptic Disorders


From here to epilepsy: the risk of seizure in patients with Alzheimer's disease Volume 18, issue 1, March 2016


The continuous aging of the population in the western world leads to an increased occurrence of neurodegenerative diseases and seizures (Cloyd et al., 2006). As the main cause of dementia (Reitz et al., 2011), Alzheimer's disease (AD) has been studied more extensively, and researchers are realising its complex relationship with seizures and epilepsy. According to the diagnostic criteria of Dubois et al. (2007), probable AD is defined as the presence of a gradual progressive and objective impairment of episodic memory, isolated or in association with other cognitive changes (e.g. visuo-spatial, language, or executive dysfunction), with supportive features such as medial temporal lobe atrophy, abnormal cerebrospinal fluid biomarkers, a specific pattern on PET imaging, or proven autosomal dominant mutation within the immediate family. Definite AD requires either clinical and histopathological evidence, or clinical and causal genetic AD mutation evidence.

It is established that patients with AD have a greater risk of seizures than non-demented patients of similar age. Seizure prevalence of AD patients ranges from 1-22% according to studies (Sjogren, 1952; Hesdorffer et al., 1996; Scarmeas et al., 2009) and the incidence rate varies between 4.8 and 11.9/1,000 person-years, which is 2-6-fold higher than subjects of similar age without dementia (Imfeld et al., 2013; Cheng et al., 2015). It is unclear, however, whether seizures are promoted by direct neuronal loss or an excitatory process. Together with cerebrovascular disorders, toxic/metabolic conditions and neoplastic diseases, dementia represents a major cause of seizures in the elderly population. Due to the impact of seizures on cognitive performances in demented patients, it seems crucial to correctly diagnose an epileptic condition in a patient with AD and treat him/her with well-tolerated antiepileptic drugs (AEDs). In this review, we will focus on the epidemiology of seizures and epilepsy exclusively in sporadic and genetic AD patients, and address the pathophysiological mechanisms underlying seizure generation in this population. We will then assess seizure types and EEG abnormalities, and finally discuss treatment strategies.


Literature search and selection criteria

To address the epidemiological aspects of seizures in patients with AD, we performed a web search on Pubmed, Medline, Scopus and Web of Science, including all studies published until September 2015 using the keywords “incidence”, “prevalence” or “frequency”, associated with “seizures” and “Alzheimer's disease”. We thus selected retrospective and prospective studies with epidemiological data on seizures in human AD patients. We also looked for references cited in review articles treating this topic. In line with the International League against Epilepsy (ILAE), we used the latest definition of epileptic seizure, i.e. sudden and transient signs or symptoms related to an abnormal and excessive discharge of neurons. Accordingly, we considered epilepsy as a chronic medical condition, characterised by an enduring predisposition to epileptic seizures. The current definition requires the occurrence of at least one seizure (Fisher et al., 2014).

In addition, we searched for publications about epilepsy treatment strategies in this particular population, using the key words “antiepileptic drugs”, “seizure treatment” and “Alzheimer”.


Epidemiology of seizures in patients with Alzheimer's disease

The major aetiologies of seizures in the elderly are cerebrovascular diseases, toxic/metabolic conditions, major neurocognitive disorders, and tumours (Hauser et al., 1996).

Since the 50s, an association between seizures and sporadic forms of AD has been observed. Sjogren et al. (1952) noticed a seizure prevalence of 22% in a small sample of pathologically-proven late-stage AD patients. Hauser et al. (1986) and Mendez et al. (1994) observed a prevalence of 9.6% and 17%, respectively, in larger cohorts of AD patients. The low prevalence rate was probably due to seizure assessment that included only convulsive episodes or was based on historical information and family questionnaires. In a population-based case-control study on first unprovoked seizures, an association with AD was found in 11% of patients (17/145), leading to a six-fold increase compared to patients without dementia (Hesdorffer et al., 1996). Sherzai et al. (2014) found a 6.9% AD prevalence among patients hospitalised for seizures in a nationwide US sample, with an odds ratio (OR) of 3.07 (95% CI: 2.98-3.16) compared to the healthy elderly population, which is higher than that for other types of dementias (OR: 2.21; 95%CI: 2.14-2.27). Recent studies have established an incidence rate (IR) of seizures in AD between 4.8 and 11.9/1,000 person-years (Irizarry et al., 2012; Imfeld et al., 2013; Cheng et al., 2015; Cook et al., 2015). Imfeld et al. (2013) observed an IR of seizures of 5.6/1,000 person-years in AD patients, while the IR was 0.8/1,000 person-years in non-demented patients. Cheng et al. (2015) found an IR of 11.9/1,000 person-years for AD patients and 5.7/1,000 person-years in control subjects. Epidemiological studies on prevalence/incidence of seizures in AD patients are summarised in table 1.

The estimation of seizure prevalence is a complex exercise, especially in patients with AD. On one hand, sudden cognitive impairment of demented patients may mislead clinicians to consider a diagnosis of seizure. On the other hand, fluctuations of cognition can be mistakenly interpreted as a seizure. As AD patients suffer from memory impairment, they might have more difficulties in reporting to their caregiver when experiencing an unusual episode. As we will see later, the majority of seizures in AD patients are focal with secondary impairment of consciousness, which may prove even more difficult to identify than generalised seizures. Moreover, seizures themselves worsen cognition at older age, especially if they are not controlled and remain unrecognised (Noebels, 2011).

There is conflicting evidence about whether AD severity is associated with a higher risk of developing seizures (Romanelli et al., 1990) or that increased risk of seizures is independent of disease stage (Amatniek et al., 2006). In a retrospective record-based study of AD patients followed in a memory clinic in Italy from 2001 to 2006, Bernardi et al. (2010) found that 9.7% had at least one unprovoked seizure in a total of 145 patients. Mean duration of the degenerative process was 3.6±1.6 years since the first seizure, and no association was established with severity of dementia. However, in a large cohort study from pooled, prospectively obtained clinical trials with a total of 3,078 patients, a higher risk of seizures with more severe AD was found (Irizarry et al., 2012). Patients with a Mini-Mental State Examination (MMSE) of 18. Seizure frequency is probably underestimated in retrospective data collection because doctors, caregivers and patients may not recollect the presence of seizures, in particular non-generalised seizures, calling for larger prospective data. The higher risk of being more severely affected is due either to neuroanatomical changes or better surveillance by experienced staff, which likely leads to identification of more possible focal seizures.

In addition, it is unclear whether young age at AD onset for sporadic cases is associated with a higher risk of seizures. Some authors suggested that this might be the case; for example, in the study of Amatniek et al. (2006), a prospective follow-up of patients with mild AD with six-month intervals showed a cumulative incidence of unprovoked seizures of nearly 8% at seven years. Younger age at onset of AD (50-59 years) was found to be an independent risk factor, with an 87-fold increase compared to the general population. A younger age at dementia onset has also been considered as a significant predictor (HR: 0.79; 95%CI: 0.70-0.90) of seizures in Cox proportional hazard ratio regression models (Irizarry et al., 2012). In contrast, Mendez et al. (1994) found that AD patients with seizures had a younger disease onset, but that epilepsy began at an advanced stage of AD (mean: 6.8 years).

It is established that patients with rare autosomal dominant forms of AD, such as mutation of presenilin-1 (PS1), presenilin-2 (PS2) and amyloid precursor protein (APP), carry a significant risk of seizures (Cabrejo et al., 2006; Larner and Doran, 2006). In a phenotypic study of five families with APP duplication, Cabrejo et al. (2006) found that 57% of patients had at least one seizure. Of the PS1 mutations, 20% were reported to be linked to seizures, which may occur either early or late during the course of the disease (Larner, 2011). Determination of seizure types can be difficult and seizures may be mistaken for “confusional episodes”.

In patients with Down syndrome, a bimodal distribution of seizures has been observed, first in the first two decades, then in the fifth and sixth decades, usually in association with the onset of dementia (Menéndez, 2005). In this context, up to 84% develop seizures (Lai and Williams, 1989). It appears that a three-phase time course occurs during the second peak, first with onset of dementia, then with progressive epileptic myoclonus, and later with non-epileptic myoclonus. The presence of APOE ε3/ε3 homozygosity is correlated with this clinical picture (d’Orsi and Specchio, 2014). Cognitive outcome of Down syndrome patients with AD has been retrospectively studied and is considered to be worse whenever associated with seizures (Lott et al., 2012). However, due to the observational design of the studies, it is not clear if seizures lead to a steeper cognitive decline in Down syndrome patients with AD or if epilepsy and dementia are both part of a more severe phenotype of Down syndrome.


Since a higher incidence of seizures among patients with AD has been established, it would be useful to understand the reasons for such an elevated susceptibility. Considering that patients with genetic forms of AD (APP, PS1 and PS2) are particularly prone to seizures, a link between amyloid-beta (Ab) and epileptogenicity has been obviously evoked. Mouse model studies may give us more insight into the putative role of Ab on generation of epileptic seizures. Palop et al. (2007) showed that accumulation of Ab in the brain could cause epileptiform activity, suggesting a direct excitatory effect of Ab on brain networks rather than a neurodegenerative mechanism. In fact, it is noteworthy to consider that AD mouse models with seizures exhibit little neuronal loss. This finding suggests, at least in experimental studies, that seizures are not related to end-stage degeneration (Chin, 2011). A similar theory is debated in humans - do patients with AD have a higher risk of seizure at the beginning of their neurodegenerative disorder or mainly at the last stage of disease?

It is suggested that Ab-induced aberrant networks are associated with sprouting of inhibitory neurons in the dentate gyrus of the hippocampus (Palop and Mucke, 2009). In addition, acute exposure to soluble peptides of Ab in hippocampal CA1 cells in mice has been shown to alter intrinsic excitability towards hyperexcitability patterns (Tamagnini et al., 2015). Mice with over-expression of APP also show an increased incidence of seizures and EEG abnormalities, such as sharp wave discharges (Palop et al., 2007; Minkeviciene et al., 2009; Born et al., 2014). In vitro experiments with mouse models showed that L-type calcium channel currents (involved in synchronous calcium oscillations) are increased with expression of human APP, independently of Ab production, suggesting a prominent role of APP in neuronal network imbalance (Santos et al., 2009).

Another argument is provided by studies on calbindin, an intracellular protein expressed in granule cells of the dentate gyrus. These cells are excitatory neurons that project information from the entorhinal cortex to mainly the CA3 portion of the hippocampus (Abraham et al., 2009). Calbindin, however, is diffusely expressed in all subfields of the hippocampus and plays a major role in long-term potentiation, plasticity and memory consolidation via its calcium binding domains. In normal aging, calbindin expression in cortical areas is globally reduced, while calbindin mRNA remains stable. In contrast, calbindin mRNA decreases in neurodegenerative conditions such as AD, Huntington's disease, Creutzfeldt-Jakob disease (CJD), and fronto-temporal lobe dementia (FTLD) (Iacopino and Christakos, 1990). Stefanits et al. (2014) showed a specific loss of calbindin immunoreactivity in dentate granule cells in late stages of AD, as compared to CJD and FTLD.

The particular relationship between Ab and seizures is emphasized by the fact that other dementias, i.e. tauopathies and synucleopathies, are generally not associated with a higher risk of seizures compared to the age-adjusted general population. Tau protein may even play a protective role in seizure generation in several mouse models, by preventing long-term potentiation impairment and NMDA receptor dysfunction (Roberson et al., 2011a). However, another model of mutant mice with over-expression of tau protein (fronto-temporal dementia with parkinsonism; FTDP-17) alone has shown increased epileptiform activity (namely spike-waves complexes) and a lower threshold for seizures, suggesting that tau imbalance, as well as Ab, can play a role in epileptogenesis (Garcia-Cabrero et al., 2013). These findings suggest that the concomitant increase in tau and Ab, as is the case in AD, may have a synergistic effect on seizure generation and open the way to treatment against neurodegeneration (Roberson et al., 2011b). Another mechanism of seizures in AD could be the presence of microbleeds due to cerebral amyloid angiopathy (CAA) since both conditions frequently co-occur (Greenberg et al., 2010; Picco et al., 2011). Indirect evidence for the role of CAA in epilepsy has recently been suggested in a large clinical sample of patients where lobar brain microbleeds were associated with late seizures after an intracerebral haemorrhage (Rossi et al., 2013). Although the coverage of these aspects are beyond the scope of the present manuscript, it is worth considering the role of neuro-inflammation in seizure and neurodegeneration. In rodent models, it has been shown that IL-1β, IL-6 and TNF are important inflammatory mediators in seizure generation (Bartfai et al., 2007). In addition, exposure to lipopolysaccharides increases intrinsic hippocampal excitability and promotes the expression of a “danger signal” protein called high mobility group box 1 (HMGB1) (Perkins, 2007), which in turn may contribute to cell loss (Ravizza and Vezzani, 2006). Moreover, it is interesting to consider that IL-1β receptor (IL-1R1) is particularly present in the hippocampus and may modulate seizure threshold. In light of these experimental hypotheses, inflammation may represent the missing link between neurodegeneration and seizures.

Semiology of epileptic seizures and EEG activity

Diagnosing a seizure in a patient with AD can be particularly difficult because sudden episodes of worsened cognition postictally can be interpreted as an ordinary fluctuation of the AD condition. Besides, as the patient has memory impairment, it can prove difficult for him/her to remember an unusual episode, especially in the case of a non-convulsive seizure. In this regard, prospective studies of patients with AD, targeting seizure incidence, should give us more insight into the seizure types and the EEG activity of AD patients. However, these are not yet available.

Vossel et al. (2013) studied 54 patients with amnestic mild cognitive impairment (aMCI) or AD in addition to diagnosed epilepsy or with epileptiform activity on EEG. This cohort of patients had transient episodes of cognitive dysfunction and onset of cognitive decline at significantly earlier age (mean age: 64.3 vs 71.1 for patients with aMCI without epilepsy; p=0.02). It is of note that the most common type of seizure was focal with impaired consciousness, mostly of a dyscognitive subtype, and that 55% of patients had non-convulsive seizures only. Seizure onset and epilepsy diagnosis preceded or coincided with aMCI or AD diagnosis for 83 and 51% of subjects, respectively. In 31/54 patients, epileptiform activity was described as sharp waves or spikes, and its localization was mostly focal, predominantly in the left temporal lobe. The authors concluded that identifying these patients early and treating them with AEDs may contribute to an improvement in their clinical course. However, this could not be demonstrated because of the retrospective design and lack of serial neuropsychological testing. Belcastro et al. (2007) reported that altered consciousness, transient aphasia and confusion, suggestive of focal seizures, are frequently encountered in AD patients suffering from epilepsy. Nevertheless, in their sample of 233 AD patients initially considered for suspected seizure, a diagnosis of epilepsy was confirmed in 22% of them. Other diagnoses were transient confusional state, syncope, or a metabolic origin. Again, the majority of patients experienced focal seizures with impairment of consciousness, with or without secondary generalization. In a retrospective study including 1,738 patients from the Mayo Alzheimer's disease registry, seizure prevalence was 3.6%, with 72% of patients experiencing focal seizures with impairment of consciousness and 52% generalised tonic-clonic episodes (Rao et al., 2009). The impact of seizures on AD course remains unclear. In an early study on early-onset AD, Samson et al. (1996) observed that seizures did not influence disease duration or mortality.

Whereas myoclonic epilepsy is rare in sporadic forms of AD (Rao et al., 2009; Vossel et al., 2013), it has been described in early-onset familial forms of AD, especially in patients with PS-1 mutations (Ezquerra et al., 1999), suggesting an independent phenotype with putative association with dysregulation of calcium homeostasis (Guo et al., 1996). Myoclonic jerks often predate tonic-clonic seizures and are associated with a worse prognosis (Fox et al., 1997).

Generally, it is considered that epileptiform activity, such as spikes and sharp waves, is rather infrequent on surface EEG in patients with dementia. A systematic review of the EEG abnormalities of 1,674 patients in a memory clinic reported such EEG findings in only 3% (Liedorp et al., 2010), while these findings were encountered more often in the cohort studied by Rao et al. (2009), in which 38% of patients were reported with epileptiform discharges, mainly uni- or bi-temporal spikes or sharp waves.

Overall, there is a patent lack of studies regarding EEG patterns of seizures in patients with AD and their effect on cognition. In addition, to our knowledge, the direct effect of seizures on memory performance in demented patients has not been studied. However, AD mouse models suggest that repetitive seizures can impair spatial memory performance (Chin et al., 2005).

Transient epileptic amnesia

Transient epileptic amnesia (TEA) is a relatively rare form of temporal lobe epilepsy, characterised by witnessed recurrent episodes of transient memory impairment, during which other cognitive functions are judged to be intact. Clinical and neurophysiological evidence indicates an epileptic condition, due to the presence of oral automatisms and olfactory hallucinations, epileptiform abnormalities on EEG, and a clear-cut response to AEDs (Butler and Zeman, 2008). In contrast to transient global amnesia, TEA consists of rather short (usually less than one hour) episodes of amnesia, and their repetitive occurrence should prompt clinicians to distinguish this treatable form of TLE from transient global amnesia (Nicastro et al., 2014). In patients with frequent temporal seizures, TEA can also mimic neurodegenerative disorders such as AD. Recent research orientates towards a continuum of TEA and dementia (Rabinowicz et al., 2000). Cretin et al. (2012) assessed the possible link between TEA and AD, by studying the case of four patients initially meeting diagnostic criteria for probable or possible TEA, but with persistent memory impairment, despite a well-conducted AED treatment. MRI showed mesial temporal atrophy stage 0 to 3, according to Barkhof's criteria and CSF analysis revealed a decreased Ab42/Tau index.

Aware of the fact that TEA patients present subtle deficits, such as autobiographical memory impairment and long-time forgetting, Butler et al., (2013) examined manual volumetry and automated MRI data of 40 patients with TEA and noticed atrophy in the bilateral hippocampal region, as well as in perirhinal and orbitofrontal cortices. As this particular form of TLE has only been recently described, follow-up studies are currently lacking to understand whether the atrophy is a cause or a consequence of TEA and whether there is an underlying, less severe, degenerative process. TEA has a relatively benign course; seizure control by AEDs is obtained in 88% of TEA subjects (Butler and Zeman, 2008). However, outcome of memory impairment is less clear.

Antiepileptic drugs in patients with Alzheimer's disease

It seems obvious that dementia patients should be treated with AEDs, once an epileptic condition is assessed. However, AD subjects represent a fragile population, as they generally are of advanced age and may suffer more frequently from AED adverse effects, including (additional) drug-induced cognitive impairment or fatigue. Tolerability and efficacy of AEDs in patients with AD have been mainly extrapolated from elderly patients without dementia. In this regard, the following studies are of particular interest. In an international multicentre randomised double-blind controlled study (Saetre et al., 2007), a similar proportion of >65-year-old patients were seizure-free with lamotrigine (LTG) and carbamazepine (CBZ) (52% and 57%, respectively; p=0.45; OR: 0.8; 95% CI: 0.45-1.43). In a randomised, double-blind trial comparing levetiracetam (LEV), LTG and CBZ, patients aged >60 years old with new-onset epilepsy showed a similar rate of seizure freedom across all three groups, but a highest retention rate for LEV at 58 weeks (61.5%; LTG: 55.6%; CBZ: 45.8%; the difference between LEV and CBZ being significant; p=0.02) (Werhahn et al., 2015).

Research of the literature on specific AED treatment strategies in AD patients is scarce. Up to now, only one randomised interventional trial has been reported. Cumbo and Ligori (2010) led a prospective, randomised, three-arm parallel-group study in patients with seizures and mild-to-moderate AD to determine the efficacy of LEV, LTG and phenobarbital (PB). AEDs were titrated to a dose with weekly increases to the following average dosages: LEV monotherapy was 956 mg (range: 500-2,000 mg/day), PB 90 mg/day (range: 50-100 mg/day), and LTG 57.5 mg/day (range: 25-100 mg/day). Efficacy was determined during a 12-month evaluation phase. While seizure control was similar across the three groups, despite low dosages with LTG, not surprisingly patients with PB suffered from more cognitive side effects. In contrast, patients who received LEV showed better attention and oral fluency performances, but mood was better with prescription of LTG. Tolerability of these three AEDs was assessed and compared with a seizure-free age-adjusted control group. Details of the results are available in table 2.

Three other studies addressed drug management of seizures in AD patients, but with an observational design. In their retrospective study including a cohort of 54 patients with mild cognitive impairment (MCI) or AD associated with seizures or EEG epileptiform activity, Vossel et al. (2013) studied the clinical response to LEV (daily dose range: 250-3,000 mg), LTG (50-600 mg), valproic acid (VPA: 250-1,500 mg), and phenytoin (PHT: 100-600 mg). The efficacy of LTG and LEV was significantly higher than that of VPA and PHT. With LTG, 53% of patients were seizure-free, while 41% were partial responders. With LEV, the respective outcomes were 44% and 50%, whereas the proportion of seizure-free patients was lower with VPA (11%, but 67% partial responders) and PHT (17% seizure-free patients). However, the daily dose range of VPA can be considered to be lower than that for LTG and LEV. LEV, LTG and VPA were better tolerated compared to PHT (p<0.05).

A retrospective open study (Rao et al., 2009) found a 79% rate of good responders (modified Engel classification I or II, i.e. seizure-free or >95% reduction of frequency and fewer than three seizures a year). The most frequently used AEDs were PHT (38.5%), followed by VPA (17.9%), and CBZ (15.4%).

In their prospective observational study of AD patients with seizure onset within Belcastro et al. (2007) studied the efficacy and tolerability of LEV in 25 patients with advanced AD and recurrent unprovoked seizures. Their data demonstrated that 11/25 (44%) were seizure-free with 500 mg/d, while this proportion increased to 72% (18/25) with increment to 2,000 mg/d. Also, 4/25 (18%) discontinued LEV because of adverse effects (e.g. confusional state or psychomotor agitation), and 2/25 were switched to another AED because of persistent seizures.

Overall, although only few studies have been reported, these suggest that for AD patients who suffer from seizures, newer AEDs have a more favourable side effect profile, but demonstrate a similar efficacy to classic AEDs. However, in some cases, average dosage of some drugs appears very low, and absence of seizures is questionable. Interestingly, Vossel et al. (2013) proposed long-term video-EEG and overnight EEGs with sleep EEG to obtain better information on the presence or absence of epileptogenic discharges. Finally, potential pharmacological interference with non-epileptic drugs, such as antihypertensive and anticoagulation drugs, is preferable to avoid life-threatening complications of AED therapy. It is recommended to initiate treatment with low dosage and proceed with a slow titration (“start low, go slow”). In addition, monotherapy should be considered, whenever possible (Arroyo and Kramer, 2001).


Evidence from observational studies suggests that AD (including its sporadic form) is associated with an increased risk of seizures. However, the exact prevalence is difficult to determine due to methodological issues and the difficulty to identify focal seizures with impaired consciousness in dementia patients.

A clear link has been established between epilepsy and AD, affecting particularly early-onset and more advanced AD. However, the pathophysiology of epilepsy is mainly speculative. From a clinical point of view, we need to better identify which AD patients are at risk of developing epilepsy and how to better diagnose seizures using, alongside EEG, additional biomarkers, such as structural MRI or functional neuroimaging. From a molecular standpoint, Ab as well as APP, and perhaps microbleeds secondary to CAA, appear to play a major role. It is of crucial importance to understand the role of APP and its cleavage products in neuronal network imbalance, as this may lead to development of new treatment strategies; in particular in light of the deleterious effects of seizures on pre-existing impaired cognition. In fact, it has been shown that when abnormal activity due to Ab in AD mouse models is suppressed, cognitive function can be improved (Sanchez et al., 2012; Vossel et al., 2013). Further clinical research is needed in order to evaluate whether neuronal network dysfunction in human patients with AD can be suppressed and cognitive function improved when treated with well-tolerated AEDs. Prospective studies of AD patients with long-term or overnight EEG, as well as cognitive outcome evaluation, are not currently available.

Supplementary data

Summary didactic slides are available on the website.


None of the authors have any conflict of interest to disclose.