Clinical outcomes in transient epileptic amnesia: A 10‐year follow‐up cohort study of 47 cases

Abstract Objective Transient epileptic amnesia (TEA) is a form of adult‐onset epilepsy where presenting features are well described, but little is known regarding prognosis. This study aimed to elucidate the long‐term prognosis of TEA regarding seizure control, memory, medical comorbidities, and life expectancy. Methods Up‐to‐date clinical information was collected for 47 people diagnosed with TEA who had joined the The Impairment of Memory in Epilepsy (TIME) study 10 years earlier. At entry to the study, information about comorbid conditions was systematically collected. Details regarding subsequent diagnoses, seizure activity, changes to treatment, or reports of cognitive impairment were obtained through the family doctor. The variables of interest were compared with UK population data. Results Mortality in the cohort was 21 of 47 (45%), with an average age at death of 82.5 years. Seizures remained well controlled for the majority but medications required adjustments in dose and type for some (28%). A small number (three cases) remained seizure‐free without medication. History of cardiovascular disorders was frequent (78.7%), typically involving hypertension (55.3%). Autoimmune disorders (25.5%), cancer (23.4%), and depression (21.3%) were also commonly reported. Although persisting memory problems were often noted, dementia was diagnosed in seven cases (14.9%). Life expectancy and comorbidities in TEA did not differ from available population norms. Significance Results suggest that life expectancy is not reduced in TEA. Although TEA does not appear to be a self‐limiting form of epilepsy, seizures are typically well controlled via medication. Because adjustments to medication may be required, even after long periods of stability, ongoing medical monitoring is recommended. Comorbid vascular disorders are frequent but appear similar to general population estimates. Monitoring mood may be important, given that people with chronic conditions are often vulnerable to depression. Because of persisting memory difficulties, the development of effective memory interventions for people with TEA is warranted.

people with chronic conditions are often vulnerable to depression. Because of persisting memory difficulties, the development of effective memory interventions for people with TEA is warranted.

| INTRODUCTION
Transient epileptic amnesia (TEA) is a form of adult-onset epilepsy in which brief episodes of amnesia recur in the absence of disruption to other cognitive functions. To meet diagnostic criteria, amnesic episodes must be witnessed, and there must be supporting evidence of epilepsy via other clinical features (eg, lip-smacking, presence of an aura), epileptiform abnormalities on electroencephalography (EEG), or through response to antiseizure medication. 1 Using these criteria, TEA may be distinguished from other similar conditions: transient global amnesia, 2 where typically only one, longer (eg, 4-6 h) amnesic episode occurs without evidence of epilepsy; 3 and the more broadly defined epileptic amnesic syndrome, where pronounced interictal memory deficits may arise in the context of subtle, nonamnestic temporal lobe seizures. 4 Following publication of the diagnostic criteria in 1998, ~250 cases of TEA have been reported. 2,[4][5][6][7][8][9][10][11][12][13][14][15] These studies have been important in identifying characteristic features, including male predominance, 15,16 onset in mid-late adult life (on average 57-63 years), monthly frequency of attacks, 6,15 and attacks on awakening (23% 6 to 85% 15 of cases). Additional epileptic phenomenon, such as olfactory hallucinations, 9 automatisms 5 or brief spells of unresponsiveness, are commonly experienced immediately preceding or in conjunction with these amnestic attacks. Mechanisms leading to the onset of TEA appear variable, including reports of limbic encephalitis, 17 structural abnormalities, 2,5,10 or neurodegenerative disease 18 ; however, in many cases the genesis remains unknown. Comorbidity with certain cardiovascular conditions (cardiac arrhythmia, cardiac valve disease, arterial aneurysm) may occur at a higher rate than in age-matched controls. 19 Other comorbidities reported in TEA cohort studies include autoimmune conditions and depression, 6 although how these rates compare with population norms has yet to be established.
Treatment for TEA typically involves a low-dose antiseizure medication, resulting in a high rate of seizure cessation. 6,13,14,19,20 Although this addresses the ictal memory disturbance, many group studies comparing patients on treatment with matched healthy controls suggest ongoing cognitive sequelae in the form of patchy retrograde amnesia for salient life events, 8,21 accelerated forgetting of recently acquired memories, [22][23][24][25] and impairment of topographical memory. 5 Standard neuropsychological tests of anterograde memory tend not to reveal severe impairments, with reductions either in the mild to moderate range 5,26,27 or in some cases, normal performance. 1,10,14,28 Other neuropsychological domains tend to remain intact. 5,6 Although these characteristics of TEA are now well documented, the long-term prognosis, with regard to seizure control, dementia risk, medical comorbidity, and seizure control has not yet been well studied. Short-term follow-up studies over a 6-to 18-month period suggest ongoing seizure control and the potential for improvements in some aspects of memory once seizures have ceased 6,9,14,27 (although in some cases ongoing memory difficulty is reported despite seizure cessation). 8,18,25 The patchy loss of memory for personal life events, however, appears to persist, 8,9,29 with only one case to date reporting recovery of some remote episodic memories that had been inaccessible. 30 Follow-up information at 5 years or beyond is extremely limited, with mixed results. In one case, a man with untreated TEA experienced ongoing seizures (~12 over the 5-year period), but maintained average performance on standard memory measures, with no evidence of general cognitive decline, and no further medical conditions reported. 29 In contrast, a woman with a 16-year history of untreated TEA subsequently developed Alzheimer's disease (AD). 31 In this case, the woman initially experienced three long episodes of amnesia over the first 6 years, but then began experiencing shorter, monthly attacks for ~10 years. Once antiseizure medication was commenced, she remained seizure-free (4 years to the end of the study), but continued to experience memory decline and met criteria for AD. The authors suggest that TEA may have been the first sign of this disease. Indeed, recent evidence from mouse models and human studies highlight the cooccurrence of both convulsive and nonconvulsive seizures with AD, particularly in cases of familial disease. [32][33][34][35] In contrast, in a 20-year follow-up of nine cases, no indication of increased risk of dementia or other negative impact on life expectancy was found. 36 However, neuropsychological re-assessment after 10 years (five people) 37 and 20 years (three people) revealed persistent memory difficulty, and in some cases, decline. With respect to epilepsy outcomes, some cases remained seizure-free, whereas others required adjustments to medication after a period of stability (a finding reported in other cases at shorter intervals 12 ). These variable accounts across a small number of people make it difficult to answer questions currently posed by the TEA community regarding whether continuous treatment is required, whether other medical conditions may be associated comorbidities, and whether further memory declines and/or dementia may be expected.
The current study provides the largest prognostic study to-date, reporting on clinical outcomes in 47 people diagnosed with TEA, over a 10-year period. In particular, this study aimed to provide information on: 1. Mortality and whether life expectancy with TEA differs from population norms; 2. The course of the epilepsy over the 10-year period (eg, seizure activity and use of medication) to provide any indications as to whether this type of epilepsy is self-limiting; 3. The prognosis with regard to dementia; 4. The prevalence of additional medical conditions with a focus on cardiovascular diseases and whether this differs from population norms.

| Participants
All 47 TEA participants were recruited from The Impairment of Memory in Epilepsy (TIME) database and had formed part of an initial series by Butler and colleagues. 5 As part of entry to the original study, the diagnosis of TEA was confirmed in all cases by one neurologist (CB), with evidence of epilepsy through any combination of epileptiform abnormalities on EEG, reports of concurrent classically epileptic features (eg, olfactory hallucinations, automatisms), and/or a positive treatment response to antiseizure therapy. The original study and follow-up study were approved by the Multicentre Research Ethics Committee, United Kingdom (MREC 03/10/77). All participants gave written, informed consent for review of their medical record.

| Procedure
Data regarding each participant's sex, date of birth, initial antiseizure medication (type and dose), and comorbidities at the time of the original study were extracted from the TIME database. To gain updated medical information, a request letter was sent to each participant's general practitioner (family doctor), asking for relevant consultation records for the 10-year period following the initial study, with a particular focus on memory complaints, ongoing seizures, and other significant medical events. Copies of any correspondence with secondary care (eg, discharge summaries) or other medical specialists (eg, neurology, psychiatry/psychology, cardiology, oncology) were also requested and reviewed by experienced medically trained members of the team (AZ and JB). For cases where medical files could not be accessed due to patient death, updated information was extracted from death certificates. Patient research files were also reviewed for any additional correspondence regarding medical updates between the research team and the families over the 10year period. Although the initial period of follow-up was between 2004/5 and 2015, where additional information regarding deaths became available during our analyses, these updates were included (ie, extending the period of follow-up to 13 years for some participants).
To gain detailed information on cognitive outcomes, all participants were invited for a neuropsychological review, repeating the same battery of tests from the original study. 5 Full results of the 14 people who were able to take part have been reported elsewhere. 37 We summarize relevant information regarding cognitive status from this study below.

| Data review
For each participant, a record was made of any additional diagnoses, seizure activity, or changes to medication, or reports of cognitive impairment (based upon GP reports, patient self-report, and/or from neuropsychological review, where available). Combined with the original information, all existing and historic diagnoses for an individual were then categorized into the following major classes of medical disorder: vascular, neurological, psychiatric, malignancy/pre-malignancy, immune-mediated, gastrointestinal, and endocrine.
Frequencies of disorders of interest (ie, vascular, immune-mediated, psychiatric, neurological, autoimmune, and malignancy) were calculated for the cohort and then considered in the context of UK population prevalence data obtained through: The Health Survey for England (2017), 38 The Alzheimer's Society Dementia UK Report 39 (the British Heart Foundation 40 ), The National Institute for Health and Care Excellence 41,42 ), Cancer Research UK, 43 and The Migraine Trust 44 ). These sources were chosen as they provide open access, UK-wide prevalence data regarding commonly occurring conditions that were reported in our population. As many of the conditions (including atrial fibrillation, stroke, and dementia) demonstrate an increasing prevalence with age, and age at the time of data collection varied extensively in the TIME cohort, population figures were obtained for multiple age ranges from the reference populations, where available.
To gain a rudimentary indication of whether cardiovascular conditions appeared more prevalent in patients with TEA than in the population at large, chi-square analyses (using p < .05 as a threshold for significance) were conducted using available population norms. Given the limited published data regarding sample size in other clinical population cohorts, it was not possible to conduct similar analyses for other comorbid conditions. In these cases, population figures are provided for reference only.

| RESULTS
Updated medical records or summary reports from GPs were obtained for 37 of the original 50 patients. In 10 additional cases, where it was not possible to locate the GP or medical record, information regarding mortality, additional diagnoses, and/or other relevant changes was obtained from the patient's family, the patient themselves, or clinical correspondence.
In the remaining three cases (two male, one female), participants were lost to contact (having moved and changed GP), and it was not possible to obtain any information regarding updated medical history or current status. As a result, these cases were removed from the analysis.
The current study is based, therefore, on 32 men and 15 women with TEA, ranging in age from 60 to 92 years (Appendix 1 for summary information on each individual case). Mean age at the time of data collection (using age at death for those deceased) was 80.2 years (standard deviation [SD] = 7.4 years).

| Mortality
Mortality in the cohort was 21 of 47 (45%), with an average age at death of 82.5 years (SD = 6.6 years; 4 female, 17 male). UK population norms, from the Office for National Statistics, 45 indicate that an average age at death of 81.3 years would be expected. Overall, pneumonia was the most common cause of death (n = 7 of the 21 patients). Other causes of death included cancer (n = 5; esophageal cancer, leukaemia, metastatic bronchial carcinoma, mesothelioma, Hodgkin's lymphoma), cardiovascular disorders (pulmonary embolus, ischaemic stroke, cardiac failure), end-stage myelofibrosis, emphysematous cystitis, and AD. For six patients the cause of death was unclear from the medical information obtained.
A comparison of the characteristics and outcomes of participants within the cohort by mortality outcome showed that those who now deceased were significantly older than the remaining participants, both at the time of TEA onset (t(45) = 2.216, p = .03) and at the time of the original study (t(45) = 3.885, p < .01). There was no indication that those who were deceased were more likely to have experienced additional seizures (χ 2 (1, N = 47) = 1.501; p = .22). As expected with increasing age and in the context of a lifelimiting condition, a higher prevalence of dementia was observed in those who were deceased (χ 2 (1, N = 47) = 5.832; p = .02). Although there was some tendency toward a higher frequency of vascular conditions in the deceased, the difference in proportion did not reach statistical significance (χ 2 (1, N = 47) = 3.189; p = .07) ( Table 1).

| Seizures and associated treatment
Information regarding medication use and any changes to dose or type was available in 36 participants. Approximately half (19/36) had remained on the same medication and dosage since the original study and had not experienced any further TEA episodes while on treatment (in one case, a brief period off treatment had resulted in two to three episodes, which ceased once medication was recommenced; another case had ceased medication for 8 years and then recommenced after seizures returned). An additional four people had made changes in their medication (either regarding dose or type of antiseizure), as a result of side effects (eg, tremor or fatigue) or concerns regarding interactions with other medications, but they had remained seizure-free on treatment. A further three had tapered off and then ceased taking any antiseizure treatments and reported no further TEA attacks.
In contrast, 10 of the 36 (28%; 6 male, 4 female), had experienced some re-emergence of seizures (in some cases these involved other seizure types such as focal impaired awareness seizure seizure or tonic-clonic seizures) while on treatment. In six cases, these attacks emerged after a long period of stability, but resolved with an increase or change in medication type, whereas in three cases, multiple changes to the treatment were required (for one participant such changes were still ongoing at the time of interview). In the final case, the patient had chosen to cease medication 3 years earlier and reported subsequently experiencing approximately two to three episodes per year.
Monotherapy was the most common approach to treatment, with only two people initially prescribed two antiseizure medications; at follow-up this increased to three people, one of whom was now on three medications (but had temporal lobe epilepsy preceding the onset of TEA). The drugs used as monotherapy at follow-up (or based on last available data) were carbamazepine (n = 12), lamotrigine (n = 9), sodium valproate (n = 6), levetiracetam (n = 6), and phenytoin (n = 3). Combinations used in polypharmacy were sodium valproate and lamotrigine (n = 2) and carbamazepine, phenytoin, and levetiracetam (n = 1).

| Medical comorbidities
Additional diagnoses were collated, with frequencies of each major medical condition class summarized in Table 2 and reference to population norms provided in Table 3.
The most common comorbidity with TEA was cardiovascular. Thirty-seven of the 47 patients with TEA (79.%; 81% of men, 73% of women) had a documented history. Hypertension was the most commonly recorded condition, reported in 55% of patients (26/47; 6 women). Ischemic heart disease (myocardial infarction or angina) was noted in 10 of the 47 records (10/47, 21%). To facilitate comparison with data available through the NHS Health Survey for England, we divided our cohort into those younger than 75 years of age at the time of data collection (n = 12), and those 75 of years of age and over (n = 35) ( Table 3). Although a numerically higher proportion of people with TEA reported comorbid ischemic heart disease (IHD) and a lower proportion of people with TEA reported comorbid hypertension, statistical comparisons with population figures indicated no significant difference (younger than 75 years IHD: χ 2 (1, 1251) = 0.392; p = .53; 75 years of age and over IHD: χ 2 (1, 1026) = 1.182; p = .28; younger than 75 years of age hypertension: χ 2 (1, 885) = 0.001; p = .98; 75 years of age and over hypertension: χ 2 (1, 531) = 0.719; p = .41).
Dementia was documented in 15% of the TEA patients (7/47, all men), a rate falling between the population estimates reported in individuals 80-84 years of age and 85-89 years of age, as seen in Table 3. In our cohort, age at dementia diagnosis was documented in five of seven cases (ages 72, 75, 76, 86, and 87; mean age of 79 years). The clinical phenotype of the dementia was noted to be consistent with AD in four patients, vascular dementia in two patients, and was unspecified in the remaining patient. Self-reported decline in memory performance for which no formal diagnosis was given was found in a further nine participants (four women, five men). Seven of these participants were formally reassessed by the TIME project, 37 with four found to have some objective evidence of decline on testing. 1 In an additional four people, ongoing, stable difficulties with autobiographical memory and accelerated long-term forgetting, or "short-term memory" were recorded, based on self-report or family report (with one case able to be confirmed via the completion of neuropsychological assessment).
Formal diagnoses of current or past major depression were noted in 21% of the cohort (10/47, 21%; three women and seven men), with two formal diagnoses of anxiety (2/47, 4%; one woman and one man). A further four participants  *Overall age range of TEA participants is 60-92 years, where n = 1 "55-64"; n = 11 "65-74"; n = 35 "75+", with mean age of 80. IHD, ischemic heart disease; AF, atrial fibrillation. 38,39,47,53 TEA, with two experiencing dual onset of TEA and depression. For the remaining four people, anxiety or depression arose after the onset of seizures. In reference to population estimates of lifetime prevalence, the rates of depression in both men and women with TEA appeared broadly similar to that of the general population (Women: TEA = 20% vs population = 23%; Men: TEA = 22% vs population = 17%). 46 Autoimmune or immune-mediated conditions were reported in 12 patients (26%). Hypothyroidism was the most common (4/47, 9%), similar to UK population rates reported in recent National Institute for Health and Care Excellence (NICE) guidance, where the prevalence of hypothyroidism in the population over 60 years of age is reported as greater than 5%. 35 Vitamin B12 deficiency was also reported in four patients (9%), although it was not clear in all cases whether the cause of this was an underlying immune disorder. This was below the estimated prevalence of vitamin B12 deficiency (all causes) in the UK of around 20% in people older than 60 years of age. 41 Few additional neurological conditions were reported to co-occur within our TEA cohort. A history of migraine was reported in 9% of patients (4/47). This compares with an estimate of the global prevalence of migraine provided by The Migraine Trust of 14.7%. 37 A range of different types of malignancy were seen (15/47, 32%; 5 women and 10 men), as compared with the Cancer Research UK report of lifetime risk of cancer of 49.7% in males, and 45.5% in females. Most commonly reported in TEA was prostate cancer (3/32 male participants, 9.4%), which was below population figures for lifetime risk of prostate cancer in males of 17.9%. 43 No other neurological condition was reported in more than one person.

| DISCUSSION
The current study is the first to report long-term follow-up data in a large series of patients with TEA, with information regarding broad clinical outcomes (mortality, comorbidities, including the risk of dementia) and how this form of epilepsy behaves over a 10-year period (regarding seizures and medication).
Consistent with our smaller study in nine cases, 36 the current results confirm within a larger group that life expectancy does not appear to be reduced, as compared with current population averages within the UK provided by the Office for National Statistics. 45 Differences between those living and deceased within the TEA cohort appeared largely age related (as those who are deceased were on average 9 years older when they entered the TIME project). There was no suggestion that ongoing seizure activity was more likely in those who are deceased.
Reports of seizure incidence and use of medication indicated that good seizure control was experienced by the majority when on treatment. In instances where patients ceased to use medication, half experienced a re-emergence of seizures. In addition, 10 participants experienced further seizures requiring increased doses, or changes in medication. This suggests that TEA is not a self-limiting form of epilepsy. However, some patients experienced long periods of freedom from seizures when off medication before the attacks re-emerged. The requirement for adjustment to medication in over a quarter of cases, even after long periods of stability, highlights the importance of ongoing medical monitoring of the condition.
As with the case study reported by Cretin and colleagues, 31 some individuals developed a type of dementia in the years following the diagnosis of TEA. The proportion (14.9%) of cases of dementia among patients with TEA was similar to current prevalence rates for people older than 80 years of age, 39 suggesting that there is not a strong association between dementia and TEA. Recent animal models suggest a relationship between amyloid beta accumulation and cortical excitability, 32 but our results indicate that this is not likely to be the underlying pathology in the majority of cases of TEA. Future studies examining the presence of amyloid (eg, through the use of Pittsburgh compound B positron emission tomography scans in vivo, or analysis at autopsy) or genetic markers of dementia such as apolipoprotein E (APOE) status in people with TEA may shed light on the inter-relationships between these two clinical conditions.
Persisting memory problems or declines in some forms of memory insufficient to attract a dementia diagnosis, were a common occurrence within the cohort. This is consistent with the well-documented, high prevalence of memory difficulties reported at diagnosis of TEA. 15 Although it is encouraging that the proportion diagnosed with dementia is not increased in comparison to the general population, and that previous work has shown stable cognitive function over 20 years of follow-up, 36 given that the majority continue to experience memory problems, there is a clear need to focus future research on establishing effective memory interventions for people with TEA.
Comorbid vascular disorders were frequent in people with TEA, but again the prevalence was in keeping with population statistics, available through the NHS Health Survey for England (2017). 38 Likewise, the prevalence of autoimmune conditions (hypothyroidism and vitamin B12 deficiency), malignancy, and migraine also appeared comparable to nationwide data, although the available details of open access data for these variables prevented more detailed comparison. No clear associations between TEA and other major medical conditions emerge from this study.
A lifetime history of mood disorders was documented in approximately one fifth of the cohort. This was most commonly in the form of major depression, which for half of the cases occurred prior to TEA onset. In some instances (two cases), the onset of depression occurred at the same time as the onset of TEA. Although not qualifying for a formal diagnosis, heightened anxiety was also noted to be a feature during seizures in three participants. Although our figures are similar to the reported UK-wide lifetime prevalence of depression of 19.7% in adults, 47 population estimates for depression can vary widely, with an increased risk in people with chronic health conditions (such as diabetes and rheumatoid arthritis). 46,48 Moreover, several studies have described an increased prevalence of depression in people with epilepsy. [49][50][51] In their systematic review and meta-analysis, Fiest et al. report the prevalence of depression in people with epilepsy as 24.13%. 52 One study of people with TEA reported a lifetime history of depression in 33% of their cohort, suggesting that depression frequently occurs. 6 Although the number of cases in our sample is too small to reliably compare with population rates, we do note that people with TEA often report higher levels of depressive symptoms, using self-report measures such as the Hospital Anxiety and Depression Scale. 5,6 This suggests that people with TEA, like others with chronic health conditions, may be more vulnerable to symptoms of psychological distress.
The overall sample size of 47 remains modest when compared with outcome studies in other conditions; however, these results provide an important step forward in knowledge regarding TEA over the long term. The current study is limited by the use of retrospective review of records, as missing information is always a possibility. Indeed, despite attempts to contact previous GPs when case information was missing or incomplete, we were unable to obtain any details for three of the original 50 participants. For a further 10 participants, information was obtained instead from the family and other clinical correspondence. For the majority of cases, however, comprehensive medical records were available, reducing any impact of this on our findings. Finally, our ability to compare the frequencies of comorbidities with established prevalence data was limited. Because our sample is small, it was not possible to conduct detailed subgroup analyses stratified by age at onset and sex.
Overall, the results of the current study suggest that TEA does not appear to limit life expectancy or increase the risk of dementia. However, regular review is advisable to ensure that medication continues to be effective, and to help manage any persisting memory or mood difficulties.

CONFLICT OF INTEREST
The authors declare that they have no financial or other conflicts of interest in relation to this research and its publication.

ORCID
Sharon A. Savage https://orcid. org/0000-0002-6477-5222 ENDNOTE 1 Evidence of decline was determined through the calculation of z-scores, comparing an individual patient's test scores with a healthy control group. Negative z-scores of −1.5 or more were interpreted as an impairment, with decline indicated by a larger negative value over time, or the appearance of a −1.5 or greater z-score that had not previously been observed on assessment.