Abstract
Purpose:
This study aims to characterize the natural history of generalized motor seizures through longitudinal stratification of patient-reported clinical seizures into high, medium and low rates of generalized motor seizures (also known as generalized tonic-clonic seizures or GTCs).
Methods:
From 2007–2018, 1.4 million seizures were recorded by 12,402 SeizureTracker.com users that met inclusion/exclusion criteria. The number of GTCs per year since the first seizure diary entry was calculated for each user and categorized as: Low (0 GTCs/year), Medium (1–2 GTCs/year), or High (≥ 3 GTCs/year) GTC rates.
Results:
Kaplan-Meier survival curves for the time until exiting the initial category were computed. There was a global difference between risk groups (p<0.01). Further pairwise log rank tests revealed a difference between each pair of risk groups (p<0.01). At 3 years, 40.8% of people initially presenting with high GTC rates remained in their initial category, while 77.3% of people initially presenting with low GTC rates remained in their initial category.
Conclusion:
A patient with a low rate of GTCs is likely to remain at low risk for future GTCs, whereas higher GTC rate patients (at least one GTC/year) may leave their initial risk stratification. Thus, yearly re-assessment may be prudent when considering risk of further GTCs. Given the association between higher yearly rates of GTCs with increased SUDEP risk and morbidity in epilepsy, further validation of these findings is important for prognostication.
Keywords: Epilepsy, Electronic Diaries, Natural History studies (prognosis), SUDEP
Introduction:
SeizureTracker.com is one of the world’s largest online and mobile patient-reported seizure diary databases (Casassa et al., 2018). It affords a unique view into self-reported seizures from a large and diverse group of patients (Ferastraoaru et al., 2018). For instance, patients self-report different seizure types, therefore it may be possible to learn what kinds of patterns are seen across time with specific seizure types of interest and seizure frequency.
Active seizure frequency has been associated with greatest burden of epilepsy-related disabilities, including social stigma, mental health comorbidities, poor quality of life, economic burden, and increased mortality (Harden et al., 2007; de Boer et al., 2008; Yoon et al., 2009; de la Loge et al., 2016). Sudden Unexpected Death in Epilepsy (SUDEP) risk for adults is 1.2/1,000 patient-years(Harden et al., 2017), and can be up to 18 times greater in patients with frequent generalized seizures. A recent meta-analysis identified the following risk factors: presence of generalized tonic-clonic (GTC) seizures, frequency of GTCs, and absence of seizure freedom(Harden et al., 2017). In another meta-analysis, the greatest risk factor was presence of three or more GTCs per year(DeGiorgio et al., 2017). These risk factors have not been systematically studied on a population level to evaluate how the natural history of how an individual’s risk evolves over time.
This study aims to stratify SeizureTracker.com patient diaries into High, Medium and Low yearly rates of GTCs based on self-report. Furthermore, we evaluate how risk for SUDEP changes over time depending on initial yearly rates.
Material and Methods:
Study Population:
Deidentified and unlinked data were extracted on March 12, 2018, from SeizureTracker.com in accordance with Beth Israel Deaconess Medical Center Institutional Review Board (IRB). At the time of export, users recorded 1.6 million seizures and clinical data between December 1, 2007 and February 28, 2018. Of the documented seizures, 1.4 million were recorded by 12,402 users with a recognizable seizure description (Figure 1). Demographic information is provided in Table 1. This platform is open to general population and thus, users were not pre-screened by a neurologist.
Figure 1: Schema illustrating SeizureTracker.com data pre-processing.
From December 2007 to March 2018, 1.6 million seizures were recorded on SeizureTracker.com. Of these, 251,301 seizures were excluded as they were not properly recorded with a recognizable seizure description: “other”, “unknown”, or missing. Of the described seizures, 268,017 were labeled as generalized motor and 1,133,490 were labeled as non-motor generalized/focal seizures.
Table 1: Baseline characteristics of study sample.
Mean (SD) (for continuous variables) or percentage (for categorical variables) are shown.
| Low GTC rate (n=7348) |
Medium GTC rate (n=2196) |
High GTC rate (n=2858) |
All Users (n=12402) |
|
|---|---|---|---|---|
| Age at initial entry diary entry, in years | 20.8 (17.3) | 20.8 (15.3) | 17.0 (13.6) | 19.9 (16.3) |
| Sex, female | 53.2%1 | 51.5%2 | 49.0%3 | 51.9%4 |
| Epilepsy etiology | ||||
| Tuberous Sclerosis Complex | 4.4% | 1.8% | 2.3% | 3.5% |
| Lennox-Gastaut Syndrome | 1.9% | 2.7% | 4.6% | 2.7% |
| Dravet Syndrome | 1.1% | 4.3% | 7.3% | 3.1% |
| Down Syndrome | 0.5% | 0.2% | 0.4% | 0.4% |
| Aicardi Syndrome | 0.4% | 0.3% | 0.6% | 0.4% |
| Rett Syndrome | 0.4% | 0.3% | 1.0% | 0.5% |
| Angelman’s Syndrome | 0.3% | 0.2% | 0.3% | 0.3% |
| Sturge-Weber Syndrome | 0.3% | 0.2% | 0.2% | 0.2% |
| Neurofibromatosis | 0.2% | 0% | 0.1% | 0.1% |
| Hypothalamic Hamartoma | 0.08% | 0% | 0.1% | 0.07% |
| Phelan-McDermid Syndrome | 0.04% | 0.1% | 0.04% | 0.06% |
| CNS tumors | 4.1% | 2.3% | 1.8% | 1.8% |
| Traumatic Brain Injury | 9.6% | 8.4% | 7.2% | 7.2% |
| CNS Infection | 4.7% | 3.9% | 4.6% | 4.6% |
| Stroke | 3.2% | 2.1% | 1.9% | 1.9% |
| Perinatal hypoxia | 3.2% | 3.6% | 2.6% | 2.6% |
| Maternal drug or alcohol abuse during pregnancy | 0.5% | 0.5% | 0.4% | 0.4% |
| Alcohol or drug abuse | 0.6% | 0.6% | 0.3% | 0.3% |
| High fever | 2.1% | 2.3% | 1.7% | 1.7% |
| CNS malformation | 5.0% | 3.1% | 4.3% | 4.3% |
| Other | 6.6% | 6.2% | 7.7% | 6.8% |
| Seizure type | ||||
| Simple partial | 19.1% | |||
| Generalized tonic clonic | 17.7% | |||
| Complex partial | 16.7% | |||
| Tonic | 16.4% | |||
| Unknown | 5.5% | |||
| Absence | 4.4% | |||
| Myoclonic | 4.7% | |||
| Other | 2.9% | |||
| Atonic | 2.7% | |||
| Secondary generalized | 2.4% | |||
| Aura only | 2.1% | |||
| Myoclonic cluster | 1.9% | |||
| Infantile spasms | 1.8% | |||
| Atypical absence | 0.8% | |||
| Clonic | 0.6% | |||
| Gelastic | 0.1% | |||
| Seizure frequency, yearly5 | ||||
| Generalized motor seizures | 0.01 (0.4) | 2.0 (2.9) | 306.7 (1677) | 148.7 (1004.1) |
| All seizures | 107 (663.7) | 82.1 (718.8) | 90.2 (388.3) | 21.2 (190.2) |
| Circadian timing of generalized motor seizures5 | ||||
| Nocturnal (7p-7a) | 52.0% | 45.6% | 46.7% | 46.7% |
| Waking (7a-7p) | 48.0% | 54.4% | 53.3% | 53.3% |
42.8% male; 4.0% not reported
44.7% male; 3.8% not reported
48.9% male; 2.1% not reported
44.5% male; 3.5% not reported
Over first year of recording
Data pre-processing:
Seizure descriptions recorded by SeizureTracker.com are shown in Table 2. Seizures descriptions were grouped into generalized motor seizures (“tonic-clonic” or “secondarily generalized”), generalized non-motor / focal seizures, or non-coded seizures (“other”, “unknown”, or missing). Inclusion criteria included seizure entries with either generalized motor or generalized non-motor / focal seizure types. Seizures with missing descriptions or recorded as “other” or “unknown” type were excluded. The number of generalized motor seizures was calculated per user for each year since the initial seizure diary entry. Based on prior literature, each year was then categorized as one of three risk stratification groups: Low (0 generalized motor seizures/year), Medium (1–2 generalized motor seizures/year), or High (≥ 3 generalized motor seizures/year)(Harden et al., 2017) The Low risk group includes users with generalized non-motor (e.g. absence) or focal seizures, but with no generalized motor seizures. For all users, for each year with no generalized motor seizures recorded, but at least 1 seizure of any kind recorded in a subsequent year, the number of generalized motor seizures was set to zero and recorded as “Low.” Thus, in order to be included in the study, users had to have documented at least one seizure of any type, but risk stratification was based on number of GTC/year, so Medium and High risk users that did not record GTC in a year but did record other seizures types were assumed to be be free of generalized motor seizures and shifted into the Low risk group.
Table 2:
Classification of SeizureTracker.Com Descriptions into Seizure Types.
| Seizure Type | Seizure Description (SeizureTracker.com) |
|---|---|
| Generalized Motor | “Tonic-Clonic”, “Secondarily generalized” |
| Non-motor Generalized/Focal | “Absence”, “Atonic”, “Atypical Absence”, “Aura only”, “Clonic”, “Complex partial”, ‘gelasdc”, “infantile spasm”, “myoclonic”, “myoclonic cluster”, “simple partial”, “tonic” |
| Non-coded events | “Other”, “Unknown”, null (not coded) |
Users have the option to describe each documented seizure, and can chose one of the following descriptions: “Absence”, “Atonic”, “Atypical Absence”, “Aura only”, “Clonic”, “Complex partial”, “gelastic”, “infantile spasm”, “myoclonic”, “myoclonic cluster”, “simple partial”, “tonic”,”tonic-clonic”, “secondarily generalized”, “other”, “unknown”; or they can choose not to describe the event (i.e. null). These descriptions were grouped as one of the following types for the purposes of this study: Generalized Motor, Non-motor Generalized/Focal, or Non-coded events.
Statistical Analysis:
For each user, an initial rate category was assigned based on the first diary year. Duration of analysis was limited to the duration of the group with minimal duration of recording (3 years; medium risk group). Users were considered “stable” in their rate category for as many years as the same category remained unchanged. Low, Medium, and High rate groups were analyzed with Kaplan-Meier analysis to determine the proportion remaining within initial rate category per year. A survival trend test (global log-rank test) was used to test for an overall significant difference between groups and pairwise log-rank tests were used to test for significant differences between individual groups. Statistical significance was evaluated at the α=0.05 level, and corrected for multiple comparisons using the Bonferroni correction. Analysis was performed in Tableau (v10.5) and R (v3.1.3).
Results:
Of the 1.4 million seizures analyzed, 19% of seizures (42.8% of all users) were generalized motor seizures. The 12,402 users were divided into low (7,348 users), medium (2,196 users), and high (2,858 users) risk groups. The low-risk group had a larger number of users compared to medium-and high-risk groups. Mean age and gender distribution was similar among the three groups. In terms of congenital etiology, a higher percentage of severe childhood epilepsies, i.e. Dravet syndrome and Lennox-Gastaut syndrome was present in the high-risk group compared to low- and medium-risk groups. Additional study sample characteristics are shown in Table 1.
Kaplan-Meier survival curves for the length of time within the initial risk stratification group are provided for people at initially evaluated at high, medium and low GTC rates in Figure 2. Of the users starting at a high GTC rate, 11.9% (339 users) changed at some point to a medium GTC rate and 10.2% (292 users) changed at some point to a low GTC rate. Of the users starting at a medium GTC rate, 6.7% (148) changed at some point to a high GTC rate and 9.6% (210) changed at some point to a low GTC rate. Of the users starting at a low GTC rate, 2.0% (146) changed at some point to a medium GTC rate and 2.0% (141) changed at some point to a high GTC rate. A significant difference in time to a departure from the initial GTC risk group was identified between the three groups (p<0.01). At three years, 40.8% of people initially presenting as high-risk and 5.1% of people initially presenting as medium-risk remained in their initial stratification, while 77.3 % of people initially presenting as low-risk remained in their initial stratification. Pairwise comparisons of groups revealed a significant difference between high- and low-risk groups (p<0.01), medium-and low-risk groups (p<0.01) and high- and medium-risk groups (p<0.01).
Figure 2: SUDEP Risk Categories over Time.
Kaplan-Meier survival curves for the length of time within initial GTC rate stratification group are shown for people with reported seizures at high (red), medium (blue), and low (green) GTC rates. The 95% confidence intervals are shaded in for each category. For High, Medium, and Low stratification groups, the number of people per group (N) and the percentage remaining at time (years) after initial risk stratification A statistically significant difference was identified between treatment groups, with pairwise analysis indicating differences between high/low risk groups, high/medium and between medium/low groups.
Discussion:
In this study, we investigated the natural history of GTC rates and provided quantitative estimates that may be used to prognosticate how rates change over time for people with self-reported seizures. We found that the vast majority of the people who initially presented with low GTC rates remained with a low rate, whereas people who initially presented with medium or high GTC rates departed from their initial level by changing groups. Higher GTC rates are significantly associated with higher SUDEP risk, high risk for injuries, and reduced quality of life (Harden et al., 2017; de Boer et al., 2008; de la Loge et al., 2016). If so, the present data suggests that patients’ risk factors for SUDEP should be evaluated yearly and counseled appropriately.
Usage of SeizureTracker.com data permits large-scale analysis of seizure patterns, based on one of the largest existing databases of patient-reported seizures in the world (Casassa et al., 2018; Ferastraoaru et al., 2018). This database can provide in-depth understanding of natural history, as users report demographic and clinical information, such as circadian timing of seizures, age of seizure onset, and anti-seizure medication usage (Goldenholz et al., 2018; Karoly et al., 2018)). However, the self-reported nature of this database lends itself to specific limitations, including misdiagnosis, misclassification of seizure types, miscounting, or lack of patient/caregiver recognition of seizures. Studies comparing electrographic to patient-reported seizures have also found that patient-reported seizure diaries may over- or under-estimate the number of seizures (Karoly and Goldenholz et al., 2018). For instance, in this study, all events, regardless of reported seizure duration, were included, and may include non-epileptic events. Although this decision may introduce bias in the results, this decision was made because inaccurate recording of the specific characteristics of a seizure does not invalidate the presence or absence of a seizure event. Patients may often correctly document the occurrence of seizures in their electronic diaries without entering valid or complete information about the characteristics of the seizure itself, in order to have appropriate daily/weekly/monthly counts to give to clinicians (Karoly and Goldenholz et al., 2018). Lastly, the analysis of electronic diary data but lead to sampling bias, as usage of electronic diaries requires consistent access to the internet as well as a smartphone or computer, and may thereby exclude patients of lower socio-economic status, who are at higher risk for SUDEP (Kaiboriboon et al., 2014).
Several assumptions were made for the analysis. Years where no generalized motor seizures were recorded were labeled as low-risk, which may have resulted in earlier category switching events for people initially assigned to high- and medium-risk groups. Analysis of history beyond three years past the initial diary entry was limited by the small sample size of the medium-risk group (n=4). Furthermore, since the years were determined based off of each user’s initial date of entry and due to the timing of data collection, the final year documented may include less than twelve months, potentially resulting in early censorship for medium groups.
In translating the results of this work to prognostication for SUDEP risk, there are several additional factors which ought to be considered, including nocturnal seizures (Harden et al., 2017). Lack of GPS location and polysomnographic or concurrent EEG recording makes analysis of nocturnal seizures challenging. Future work would benefit from the use of wearable biosensors that increases certainty about presence and absence of events at any time of day or night, as well as information about sleep time (Onorati et al., 2016). In addition to other risk factors, confounders were not accounted for in the survival analysis, so that differences in survival between risk groups may be partially attributable to baseline differences in demographic or seizure characteristics. For instance, there is a lower percentage of severe childhood epilepsies in the low-risk group, compared to the high-risk group. Anti-seizure medications are less reliably recorded in the SeizureTracker.com database, and therefore subgroup analysis to evaluate the association of anti-seizure medication use with natural history transitions is not explored here. However, future work may explore demographic differences, seizure characteristics, anti-seizure drugs, as well as how presence of nocturnal seizures should be best combined with GTC frequency in order to produce a more fine-grained risk stratification.
Despite these limitations, it is helpful to consider that perhaps a patient at low-risk for SUDEP is likely to remain low-risk (based on GTC rates), whereas higher risk patients (defined as at least one GTC/year) should be re-assessed for risk categorization yearly, as they are expected to change risk category over time. Additional research may be useful for examining whether similar patterns of SUDEP risk natural history are present among physician-verified seizures and among electrographic seizures.
HIGHLIGHTS.
A large scale patient-reported diary system was evaluated for changes in GTC rates across years
A patient with a low rate of GTCs is likely to remain at low risk for future GTCs
A patient with higher GTC rate (at least one GTC/year) may leave their initial risk category
It may be imported to re-asses GTC rates yearly in patients
Acknowledgements:
Use of data was facilitated by the International Seizure Diary Consortium (https://sites.google.com/site/isdchome/), and by generous assistance from SeizureTracker.com™.
Funding:
Dr. Ayub reports no disclosures.
Dr. Chiang reports no disclosures.
Mr. Moss is the founder of SeizureTracker.com
Dr. Goldenholz reports funding from NIH T32NS048005.
Study Funding: This study was funded in part by T32NS048005 Neurostatistics and Neuroepidemiology Fellowship.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Disclosures of conflicts of interest:
DG reports funding from NIH, BIDMC and is an advisor for Magic Leap.
The other authors have no conflicts of interest to disclose.
Conflict of interest
We would like to have the manuscript considered for publication in Seizure. We think this study may be of interest to your readers in terms of nature history of GTC frequency and possible implications, i.e., SUDEP risk, but did not frame the longitudinal context for that risk. Our study provides additional information about what kinds of expectations one might have about risk stratification over time, which we think may be of use to patients with epilepsy and epileptologists.
This study was presented as a poster at the Annual Epilepsy Society Conference in December 2018.
References:
- de Boer HM, Mula M and Sander JW (2008) ‘The global burden and stigma of epilepsy’, Epilepsy and Behavior, 12(4), pp. 540–546. doi: 10.1016/j.yebeh.2007.12.019. [DOI] [PubMed] [Google Scholar]
- Casassa C, Rathbun Levit E and Goldenholz DM (2018) ‘Opinion and Special Articles: Self-management in epilepsy: Web-based seizure tracking applications’, Neurology, 91(21), pp. 2027–2030. doi: 10.1212/WNL.0000000000006547. [DOI] [PMC free article] [PubMed] [Google Scholar]
- DeGiorgio CM et al. (2017) ‘Ranking the leading risk factors for sudden unexpected death in epilepsy’, Frontiers in Neurology. doi: 10.3389/fneur.2017.00473. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ferastraoaru AV et al. (2018) ‘Characteristics of large patient-reported outcomes: Where can one million seizures get us?’, Epilepsia Open, pp. 0–2. doi: 10.1002/epi4.12237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Goldenholz DM et al. (2018) ‘Different as night and day: Patterns of isolated seizures, clusters, and status epilepticus’, Epilepsia, 59(5), pp. 73–77. doi: 10.1111/epi.14076. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Harden C, Tomson T, Gloss D, Buchhalter J, Cross JH, Donner E, French Jacqueline A, et al. (2017) ‘Practice guideline summary: Sudden unexpected death in epilepsy incidence rates and risk factors: Report of the Guideline Development, Dissemination, and Implementation Subcommittee of the American Academy of Neurology and the American Epilepsy Society’, Neurology, 88(17), pp. 1674–1680. doi: 10.1212/WNL.0000000000003685. [DOI] [PubMed] [Google Scholar]
- Harden CL et al. (2007) ‘The effect of seizure severity on quality of life in epilepsy’, Epilepsy and Behavior, 11, pp. 208–211. doi: 10.1016/j.yebeh.2007.05.002. [DOI] [PubMed] [Google Scholar]
- Kaiboriboon K et al. (2014) ‘Premature mortality in poor health and low income adults with epilepsy.’, Epilepsia. United States, 55(11), pp. 1781–1788. doi: 10.1111/epi.12789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Karoly P, Goldenholz DM and Cook M (2018) ‘Are the days of counting seizures numbered?’, Current Opinion in Neurology. doi: 10.1097/WCO.0000000000000533. [DOI] [PubMed] [Google Scholar]
- Karoly PJ et al. (2018) ‘Circadian and circaseptan rhythms in human epilepsy: a retrospective cohort study’, The Lancet Neurology, 17(11), pp. 977–985. doi: 10.1016/S1474-4422(18)30274-6. [DOI] [PubMed] [Google Scholar]
- de la Loge C et al. (2016) ‘PatientsLikeMe® Online Epilepsy Community: Patient characteristics and predictors of poor health-related quality of life’, Epilepsy and Behavior, 63, pp. 20–28. doi: 10.1016/j.yebeh.2016.07.035. [DOI] [PubMed] [Google Scholar]
- Onorati F et al. (2016) ‘Improvement of a convulsive seizure detector relying on accelerometer and electrodermal activity collected continuously by a wristband’, Epilepsy Pipeline Conference, San Francisco, California. [Google Scholar]
- Yoon D et al. (2009) ‘Economic impact of epilepsy in the United States’, Epilepsia, 50, pp. 2186–2191. doi: 10.1111/j.1528-1167.2009.02159.x. [DOI] [PubMed] [Google Scholar]