A 2% random sample of veterans 55 years and older was generated from the Veterans Health Administration National Patient Care Database. After exclusion, a final sample size of 292 262 veterans with and without late-onset unprovoked seizures of unknown etiology (LOSU) were followed up for incident dementia.
Adjusted cumulative incidence of dementia is shown for veterans with and without LOSU at baseline accounting for the competing risk of mortality. Age is used as the time scale to indicate age at dementia diagnosis. Models were adjusted for demographic variables, cardiovascular risk factors, depression, and traumatic brain injury.
eTable 1. List of ICD-9-CM diagnosis codes used to identify prevalent and incident dementia.
eTable 2. List of ICD-9-CM diagnosis codes used to exclude secondary causes for late-onset unprovoked seizures.
eTable 3. Association of incident unprovoked seizures with risk of dementia after excluding cases with moderate, severe, and unclassified TBI (n = 2340).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Keret O, Hoang TD, Xia F, Rosen HJ, Yaffe K. Association of Late-Onset Unprovoked Seizures of Unknown Etiology With the Risk of Developing Dementia in Older Veterans. JAMA Neurol. Published online March 09, 2020. doi:10.1001/jamaneurol.2020.0187
What is the association of late-onset unprovoked seizures of unknown etiology with the risk of dementia?
In this cohort study that included 292 262 US veterans 55 years and older, those with incident unprovoked seizures of unknown etiology were twice as likely to be diagnosed as having dementia during follow-up.
Elderly individuals with new-onset unprovoked seizures of unknown etiology should be followed up for signs of dementia.
The incidence of unprovoked seizures and epilepsy increases considerably in late life, with approximately one-third of seizures being of unknown etiology. While individuals with dementia have a high risk of developing unprovoked seizures, it is unknown whether older adults with late-onset unprovoked seizures of unknown etiology (LOSU) are at risk of developing dementia.
To determine whether incident LOSU is associated with a higher risk of dementia among older US veterans.
Design, Setting, and Participants
This retrospective multicenter cohort study was conducted using data from US Veterans Health Administration medical centers from October 2001 to September 2015. Data were generated from all veteran inpatient and outpatient encounters that occurred within Veterans Health Administration facilities. A random sample of 941 524 veterans 55 years and older was generated. A total of 649 262 veterans previously diagnosed (using International Classification of Diseases, Ninth Revision, Clinical Modification codes) with dementia, unprovoked seizures, epilepsy, and conditions that could lead to seizures (brain tumors, trauma, infections, stroke, and neurotoxin exposure) as well as veterans without follow-up data were excluded. Data were analyzed from October 2018 to July 2019.
Late-onset unprovoked seizures of unknown etiology were defined as a new diagnosis of epilepsy or unprovoked seizures without a diagnosis of a secondary cause for seizures. Incident LOSU was assessed during a 5-year baseline period.
Main Outcomes and Measures
Veterans were assessed for incident dementia diagnosis during an outcome period. Fine-Gray proportional hazards models were used to determine whether LOSU was associated with greater risk of incident dementia. Models were adjusted for demographic variables, cardiovascular risk factors, depression, and traumatic brain injury.
Of the 292 262 included veterans, 282 628 (96.7%) were male, and the mean (SD) age was 73.0 [8.8] years. During the baseline period, 2166 veterans developed LOSU. The mean (SD) follow-up after LOSU was 6.1 (2.9) years. After multivariable adjustment, veterans with LOSU had greater risk of dementia compared with veterans without seizures (hazard ratio, 1.89; 95% CI, 1.62-2.20). A sensitivity analysis imposing a 2-year lag between incident LOSU and dementia diagnosis led to similar results.
Conclusions and Relevance
These findings suggest LOSU in older veterans is associated with a 2-fold risk of developing dementia. While seizures are commonly thought to occur in late stages of dementia, these findings suggest unexplained seizures in older adults may be a first sign of neurodegenerative disease.
The incidence of unprovoked seizures and epilepsy markedly rises in late life, with rates in older adults ranging from 2-fold to 6-fold higher than in younger adults.1-5 These high rates are because of age-related accumulation of seizure-related structural disease of the brain, such as cerebrovascular disease (CVD), intracranial tumors, traumatic brain injury (TBI), previous brain infections, and neurodegenerative disease.6,7 However, in approximately one-third of all new-onset unprovoked seizures, a cause is not found.8-11
Patients with dementia are also prone to develop seizures compared with age-matched healthy controls.6,12,13 Transgenic mouse models of Alzheimer disease (AD) suggest these high rates of seizures are secondary to AD neuropathology, specifically β-amyloid (Aβ) oligomeres.14 While seizures usually develop in advanced stages of AD,15 they may appear in an earlier disease stage.16,17 Recent advances in AD biomarker research have shown accumulation of Aβ to precede the onset of clinical symptoms by up to decades.18,19 In a 2019 study,20 individuals with late-onset unprovoked seizures of unknown etiology (LOSU) were shown to have similar cerebrospinal fluid Aβ 1 to 42 concentration as patients with AD. Thus, it is plausible that dementia-associated neuropathology, such as Aβ accumulation, causes seizures in some patients years before they manifest clinical signs of dementia.
Only a few studies have examined the association of seizures or epilepsy with risk of developing dementia,21-28 mostly in small sample sizes. These studies examined all-cause seizures, thereby failing to control for diseases that cause both seizures and dementia, such as CVD and TBI. Thus, the association of late-life onset seizures with onset of dementia is not clear. The objective of our study was to investigate the association of LOSU with subsequent development of dementia. We hypothesized that a diagnosis of LOSU can be used to identify individuals with a high risk of developing dementia over several years either as a prodrome or risk factor.
We conducted a retrospective cohort study. All procedures were approved by the University of California, San Francisco, institutional review boards; San Francisco Veterans Affairs Medical Center; and US Army Medical Research and Materiel Command, Office of Research Protections, Human Research Protection Office. The need for informed consent was waived because the data were deidentified administrative data.
We generated a random sample of 941 524 veterans 55 years and older from the US Veterans Health Administration (VHA) National Patient Care Database. This database contains all veteran inpatient and outpatient encounters that occur within VHA health care facilities. Veterans were included in the study from their first encounter during the study period from October 2001 to September 2015.
We excluded possible secondary causes of seizures by excluding all veterans with a previous diagnosis of CVD, intracranial tumors, previous brain infections, dementia, or neurotoxin exposure. For a list of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes used, see eTables 1 and 2 in the Supplement. Veterans with a previous diagnosis of TBI were not excluded as most cases of TBI are mild, and we took this into account in our analytic models.
Veterans’ inclusion was followed by evaluation of seizure incidence during a 5-year baseline period. This period was necessary to adequately separate veterans with and without seizures. We defined LOSU as a new diagnosis of epilepsy or seizures (ICD-9-CM codes 345.00 through 345.91 and 780.39) occurring at any time during the baseline period. These ICD-9-CM codes were previously shown to have a high predictive value for epilepsy and seizures in adults.29 As it is possible that seizures may be the presenting feature of an undiagnosed structural brain lesion, we excluded veterans diagnosed with a clear cause for seizures or dementia at any time during the baseline period, even if these were made after LOSU diagnosis. This conservative approach provided additional confidence that observed new seizures were of unknown etiology. We also excluded veterans without a follow-up evaluation after the baseline period. Our final sample size was 292 262 (Figure 1).
We evaluated veterans for incident dementia during an outcome period. This period lasted until a diagnosis of dementia, mortality, or the end of the study (September 2015). Prevalent dementia before and during the baseline period and incident dementia during the outcome period were identified with a comprehensive list of ICD-9-CM codes generated by the VHA Dementia Steering Committee in 2016.30 For a list of ICD-9-CM codes used, see eTable 1 in the Supplement.
Demographic variables included were age, sex, race/ethnicity, and socioeconomic status based on neighborhood education and income data from the 2016 American Community Survey (5-year estimates). We defined education according to veteran zip code tabulation area, where 25% or less vs more than 25% of the adult population had completed a college education (bachelor’s degree or higher). Income was defined as a 3-level variable (low, middle, high) and categorized by tertile of median zip code tabulation area income for adults. We used ICD-9-CM codes to identify medical diagnoses for diabetes, hypertension, myocardial infarction, congestive heart failure, TBI, and major depressive disorder at baseline.
Date of death was determined using the VHA Vital Status File, which combines information from the VHA, the US Center for Medicare & Medicaid Services, and the Social Security Administration to determine the date of death.31 Prior studies have found that the VHA Vital Status File is comparable with the National Death Index in accuracy and completeness.32
Incidence of LOSU at baseline was determined, and characteristics of veterans with and without LOSU at baseline were compared using χ2 test for categorical variables and t test for continuous variable. Incidence of dementia during the outcome period was determined for veterans with and without LOSU. We used a Fine-Gray regression model to examine time to dementia diagnosis in our study.33,34 The Fine-Gray regression provides a more conservative estimate of the association because it views mortality as a competing risk for dementia rather than censoring mortality, as done in Cox regression models. We first conducted analyses using models that adjusted for demographic characteristics (age, sex, education, race/ethnicity, and income) and then using models adjusted for possible confounding variables associated with dementia, including medical comorbidities, depression, and TBI.
Analyses were performed examining the primary outcome of all-cause dementia and reported as hazard ratios (HRs) with 95% CIs. Cumulative incidence of dementia adjusted for demographic characteristics and comorbid medical conditions was plotted by age in veterans with and without LOSU. All P values were 2-tailed, and significance was set at P < .05. All analyses were performed with SAS version 9.4 (SAS Institute).
Of the 292 262 included veterans, 282 628 (96.7%) were male, and the mean (SD) age was 73.0 [8.8] years. A total of 2166 veterans (0.7%) had incident LOSU during the baseline period. The mean (SD) time from inclusion to LOSU diagnosis was 2.2 (1.6) years. Veterans with LOSU were more likely to be younger, black, have lower income, and have higher baseline prevalence of comorbidities (Table).
During the outcome period, 58 608 veterans died (20.1%); veterans with LOSU were more likely to die than those without LOSU (601 of 2166 [27.7%] vs 58 007 of 290 096 [20.0%]; P < .001). Incident dementia was diagnosed in 14 076 veterans, of whom 181 (1.3%) had prior LOSU. The mean (SD) follow-up time from LOSU diagnosis until an outcome was 6.1 (2.9) years.
The unadjusted hazard of dementia accounting for the competing risk of death was almost double for veterans with LOSU than in those without LOSU (HR, 1.99; 95% CI, 1.71-2.32). The association of LOSU with dementia risk remained elevated after adjustment for demographic and medical comorbidities (adjusted HR, 1.95; 95% CI, 1.67-2.27). An additional adjustment for depression and TBI resulted in slightly attenuated but still elevated association (adjusted HR, 1.89; 95% CI, 1.62-2.20). Excluding cases of moderate and severe TBI from our analysis resulted in similar results (eTable 3 in the Supplement). The cumulative incidence of dementia adjusted for demographic characteristics and comorbid conditions in patients with and without LOSU is shown in Figure 2.
To mitigate possible reverse causation, we conducted a sensitivity analysis imposing a 2-year lag between LOSU diagnosis and dementia diagnosis. The estimated effect sizes and significance were similar to our primary model (HR, 1.71; 95% CI, 1.39-2.11).
In this cohort study of nearly 300 000 veterans, LOSU was associated with a 2-fold risk of being diagnosed as having dementia during follow-up. While veterans with LOSU were more likely to have prevalent medical comorbidities, accounting for these comorbidities resulted in similar effect size and significance. In addition, imposing a 2-year lag between LOSU and dementia diagnosis did not substantially change these findings and helps refute reverse causation.
Few prior studies have assessed the risk of dementia following seizures or epilepsy and have had mixed results. To our knowledge, only 1 study35 retrospectively assessed patients discharged from inpatient care with a diagnosis of all-cause epilepsy and reported that those with epilepsy had an increased risk of dementia. Several small case-control studies21-25 have not found an association of seizures with the risk of dementia. These mixed results are likely because of difference in study design, heterogeneity in definition of seizures, and variable sample size. To our knowledge, our study is the first nationwide cohort from both inpatient and outpatient samples excluded for secondary causes of seizures to measure the association of LOSU with risk of dementia.
We hypothesized 2 major, possibly complementary, and difficult to disentangle associations between seizures and dementia: seizures cause or accelerate neurodegeneration and neurodegeneration causes epilepsy. Several studies have demonstrated an association of seizures with subacute deterioration, faster progression,36 and earlier onset of dementia.15,27 Rapidly progressive dementias, eg, Creutzfeldt-Jakob disease, are commonly associated with seizures during or preceding cognitive decline.37 Clinicopathologic series in postmortem or surgically resected temporal lobes from patient with temporal lobe epilepsy have shown higher amounts of Aβ plaques and neurofibrillary tangles compared with controls without epilepsy.38-40 Additional supporting evidence for seizures being secondary to neurodegenerative neuropathology comes from studies of familial AD using mice with human amyloid precursor protein (J20 line). These studies suggest Aβ brain deposition impairs synaptic plasticity, which results in aberrant network excitation leading to epilepsy.41 Individual variability in synaptic remodeling as a response to neuropathology may account for differences in the time of seizure onset relative to cognitive symptoms. Genetic factors may also play a mechanistic role; both first-degree relatives of patients with AD and apolipoprotein E 4 allele carriers without dementia exhibit high-voltage slow waves and sharp waves on hyperventilation electroencephalograms,42,43 suggesting susceptibility for seizures exists in individuals at risk of developing AD.
Our study has a number of important strengths, including the large sample size, longitudinal follow-up, and conservative participant screening criteria. We applied the rigid criteria to approximate LOSU rather than all-cause seizures by excluding veterans with diagnoses that can cause both seizures and cognitive decline, both before and during baseline period. Moreover, veterans who received a dementia diagnosis during the baseline period, either before or after LOSU diagnosis, were excluded to allow group comparison to veterans without LOSU as well as to reduce the possibility of reverse causality. We also adjusted for multiple medical and demographic variables as well as competing risk of death to test the robustness of our findings.
Our study has limitations. The major limitation of our study was our reliance on ICD-9-CM codes to establish diagnoses, which are less sensitive than structured diagnosis and may be subject to miscoding. Furthermore, ICD-9-CM codes have been shown to be slightly less accurate for mild CVD,44 suggesting that some veterans defined as having LOSU had CVD. Nevertheless, most studies on poststroke epilepsy implicate moderate to severe CVD as a probable cause, whereas the evidence for mild or subcortical CVD is very limited.45,46 Data on seizure severity and subtype as well as dementia subtype was limited. Additionally, our veteran sample included mostly men, making generalizations of our results to women challenging.
In this cohort of 292 262 older veterans, LOSU diagnosis was associated with a doubling of the risk of developing dementia during follow-up and remained significant after adjusting for demographic and medical conditions. While seizures commonly occur in late stages of dementia, our results suggest it can also precede dementia by several years. These findings emphasize the need for comprehensive research on late-onset seizures and dementia. Considering the growing elderly population and high incidence of LOSU, further studies may have important ramifications for dementia research and clinical practice.
Accepted for Publication: January 16, 2020.
Corresponding Author: Ophir Keret, MD, Global Brain Health Institute, University of California, San Francisco, Sandler Neurosciences Center, 675 Nelson Rising Ln, Ste 190, San Francisco, CA 94158 (firstname.lastname@example.org).
Published Online: March 9, 2020. doi:10.1001/jamaneurol.2020.0187
Author Contributions: Dr Keret had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Keret, Yaffe.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Keret, Yaffe.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Xia.
Obtained funding: Rosen, Yaffe.
Administrative, technical, or material support: Keret, Hoang.
Study supervision: Yaffe.
Conflict of Interest Disclosures: Dr Rosen has received grants from the National Institute on Aging and Biogen as well as personal fees from Ionis Pharmaceuticals and Novartis. No other disclosures were reported.
Funding/Support: This research was supported by grant W81XWH-12-PHTBI-CENC from the US Department of Defense (Dr Yaffe), grant K24 AG031155 from the National Institute on Aging (Dr Yaffe), and the Sierra Pacific VISN Mental Illness Research, Education, and Clinical Centers (Dr Yaffe).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Create a personal account or sign in to: