Risk of Epilepsy Following a First Posttraumatic Seizure

Abstract

Background and Objectives

Traumatic brain injury (TBI) is a common cause of epilepsy, and the risk increases with injury severity. Whether a first posttraumatic seizure (PTS) represents epilepsy is a common clinical problem, but often unknown. Prognostication is important for providing correct patient information and consideration of antiseizure medication. Our objective was to understand how trauma severity and latency from the injury affect the risk of epilepsy after a first PTS.

Methods

The register-based cohort study including all individuals hospitalized following a TBI in Sweden 2000–2010, in addition to 3 age-matched and sex-matched controls per case. We analyzed the 10-year probability of epilepsy following a first seizure using the Kaplan-Meier estimator.

Results

The risk of an epilepsy diagnosis was 41.1% (95% CI 38.6–43.7) following a PTS, higher than the risk of 33.4% (95% CI 30.3–36.5) in those without prior TBI. The risk increased with injury severity, with the highest risk following focal cerebral injuries, 62.3% (95% CI 53.7–70.9). Mild injuries and skull fractures showed a similar risk to the group without previous TBI. In addition, the risk was higher if the seizure occurred <2 years following the trauma.

Discussion

Severity of the injury and latency are major modulators of epilepsy risk following a first PTS. The risk was high in the most severe types of TBI, but a substantial proportion did not develop epilepsy, highlighting the need for further research on prognostication and biomarkers, as well as caution in diagnosing epilepsy based on a first PTS.

Introduction

Traumatic brain injury (TBI) is a common cause of epilepsy.^1,2 An important clinical issue is the recurrence risk following a first posttraumatic seizure (PTS). In cases of high risk, treatment with antiseizure medications (ASM) may be warranted. The 2014 International League Against Epilepsy (ILAE) practical definition of epilepsy also states that an individual could be diagnosed with epilepsy following a single unprovoked seizure with an estimated risk of ≥60% for subsequent seizures within 10 years.³ Better understanding of when a PTS is likely to recur could therefore provide pathophysiologic insight as well as inform management and allow early diagnosis of epilepsy.

The reported recurrence risk following a first unprovoked PTS varies in the literature. In a longitudinal cohort study from 1997 on moderate to severe TBI and at least one risk factor for posttraumatic epilepsy (PTE), the risk of a subsequent seizure within 2 years was 86%.⁴ By contrast, the 10-year risk of epilepsy was 46.6% in a retrospective cohort study of moderate to severe TBI.⁵ Overall, recurrence risks range from 46 to 100%.⁶ In addition to TBI severity, timing of the first PTS could be a potential prognostic indicator. In poststroke epilepsy, another form of structural acquired epilepsy, the risk of seizure recurrence is higher during the first 2 years following the cerebral insult and then decreases with time.⁷

We used population-wide data to analyze the risk of epilepsy following a first PTS, attempting to test the hypotheses that a more severe TBI and a shorter time from the TBI would increase the risk of seizure recurrence.

Methods

We used data from the Swedish National Patient Register (NPR), the Cause of Death Register, and the National Prescribed Drug Register, all managed by the National Board of Health and Welfare, and the Population Register managed by Statistics Sweden. The data linkage was based on the personal identification number and performed by the register holders. Information from the NPR is available for in-patient care from 1987 and out-patient care from 2001, with improved coverage until 2005. The Cause of Death Register contains information regarding dates and causes of deaths from 1961. Reporting to the national registers is mandatory for all health care providers.

Cohort

In a dataset compiled as previously described, we included all individuals aged ≥18 years hospitalized for TBI in Sweden between 2000 and 2010 and 3 age-matched and sex-matched controls per case without TBI.² From this initial dataset, we identified all patients with a first seizure diagnosis >7 days (to exclude acute symptomatic seizures) following the TBI or the corresponding inclusion date in the non-TBI group. To achieve as large groups as possible, we included all individuals fulfilling the inclusion criteria (Figure 1). The first seizures were approximately three-fold more common in the TBI group, resulting in similarly sized first seizure samples in the TBI and the non-TBI group.

Figure 1. Inclusion Flowchart.

Definitions

Diagnoses in the NPR are based on the International Classification of Diseases (ICD) system. Seizure and epilepsy were defined as the ICD-10 codes R568 and G40, respectively. Trauma severity levels were defined using the following ICD-10 codes: mild—S060; fracture—S020, S021, S027, and S029; extracerebral injury—S064, S065, and S066; diffuse cerebral injury—S061, S062, S067, S068, and S069; and focal cerebral injury—S063. The comorbidities stroke, CNS tumors, and CNS infections have been described previously.²

Seizure Diagnosis Validation

Since our aim was to analyze unprovoked seizures, we excluded individuals with diagnostic codes coregistered with the seizure diagnosis that could indicate an acute symptomatic seizure.⁸ We also conducted a review using medical records from Södra Älvsborg Hospital and retrieved medical records for individuals with a first ICD code of R568 > 7 days following admission for TBI. We found the following associated ICD-10 codes to indicate acute symptomatic seizure or misdiagnosis: E106A, F102, F103, F199, F259, F329, I609, I633, I619, I691, I694, R559, S065, and Y1599. In total, individuals with the following ICD-10 codes coregistered with the seizure diagnosis were excluded as potential acute symptomatic seizures: CNS infection and inflammation: A066, A17, A39, A8, B003, B004, B010, B011, B020, B021, B050, B051, B060, B220, B261, B262, B375, B384, B431, B500, B582, B602, and G00-08; CNS tumors: C70-72 and D33; metabolic and electrolyte disturbances: E106A, E116A, E16, and E871; psychiatric causes: F1, F2, and F32; stroke: I6; other associated symptoms: R291 and R559; traumatic brain injury: S06, S020, S021, S027, and S029; and intoxication: Y1.

Statistical Analyses

Statistical analyses were performed using SPSS, version 28. The risk of epilepsy was analyzed using the Kaplan-Meier estimator. The time was calculated from first seizure diagnosis occurring >7 days following the index TBI, to the time of epilepsy diagnosis or censoring at death, >10 years from the seizure diagnosis, or on 2017-12-31. Confidence intervals of Kaplan-Meier estimates were calculated using ±1,96 x standard error of the mean. Log-rank analyses were used to analyze risk differences.

Hazard ratios (HR) were calculated using the Cox proportional hazard model, unadjusted and with adjustments for age at first seizure and sex, in addition to time-dependent adjustments from the time of the diagnosis of the comorbidities stroke, and CNS tumor and CNS infection to the first seizure. Since the risk of epilepsy was particularly high during the first and second years following the first seizure, we also performed time-stratified Cox proportional hazard models for year 1, year 2, and years 3–10 to further increase the validity of the analyses. The significance level was set at p < 0.05 in all analyses.

Sensitivity Analysis

The epilepsy diagnosis has previously been validated in Swedish registers with an accuracy >90%.⁹ The combination of an ICD-10 code of epilepsy and an ASM prescription is internationally associated with a high accuracy of the epilepsy diagnosis.¹⁰ As a sensitivity analysis testing our outcome measure, we analyzed the proportion of individuals with an epilepsy diagnosis who were also prescribed an ASM in a range of ±7 months from the epilepsy diagnosis. The time frame ±7 months was chosen because, in Sweden, prescriptions are generally valid for 12 months, allowing the inclusion of prescription renewals.

Standard Protocol Approvals, Registrations, and Patient Consents

The study was approved by the Gothenburg Regional Ethical Review Authority (approval number 612-18), with supplementary approval from the Swedish Ethical Review Authority (approval number 2020-05132). Data were anonymized by the National Board of Health and Welfare.

Data Availability

The data from the Swedish NPR cannot be shared by the authors because of confidentiality laws.

Results

We identified 4,239 individuals with a first seizure, 2,286 in the TBI group and 1,953 in the non-TBI group, translating into an odds ratio of a first seizure of 3.46. The median age at the time of the seizure was 51 (range 18–99) in the TBI group and 67 (range 19–101) in the non-TBI group. Male sex was more common in both groups. Mild injury was the most common injury type. Stroke and CNS infection were more common in the TBI group, and CNS tumor was more common in the non-TBI group (Table 1).

Table 1.

Demographics

	TBI	%	Non-TBI	%
	n		n
Age at first seizure
18–39	702	30.7	441	22.6
40–59	730	31.9	352	18.0
60–79	519	22.7	550	28.2
80+	335	14.7	610	31.2
Sex
Male	1,477	64.6	1,241	63.5
Female	809	35.4	712	36.5
Injury
Non-TBI	0	0	1953	100
Mild	1,426	62.4	0	0
Fracture	111	4.9	0	0
Extracerebral injury	308	13.5	0	0
Diffuse cerebral injury	255	11.2	0	0
Focal cerebral injury	186	8.1	0	0
Comorbidities
Stroke	517	22.6	351	18.0
CNS tumor	64	2.8	150	7.7
CNS infection	55	2.4	26	1.3

Abbreviation: TBI = traumatic brain injury.

Diagnostic Code Validation

In our review of medical records, we found a 96.5% accuracy of the seizure diagnosis. After excluding individuals with coregistered ICD-10 codes indicating acute symptomatic seizures and misdiagnosis, 26.8% of the remaining seizures were considered acute symptomatic. In addition, 7.3% of the cases had previous seizures when presenting with the first PTS, thereby fulfilling the diagnostic criteria for epilepsy.

Epilepsy

During the total follow-up period (median 2.0 years), 32.0% of the TBI group and 22.4% of the non-TBI group received an epilepsy diagnosis following their first seizure. An epilepsy diagnosis was detected in 52.2% following focal cerebral injury, 45.1% following diffuse cerebral injury, 34.4% following extracerebral injury, 31.5% following fracture, and 26.5% following mild injury. In individuals with an epilepsy diagnosis, 92% were prescribed ASM ±7 months from the time of diagnosis.

Risk of Epilepsy

The 10-year risk of an epilepsy diagnosis was 41.1% (95% CI 38.6–43.7) for patients with a first PTS. The corresponding risk in the non-TBI group was 33.4% (95% CI 30.3–36.5) (Figure 2). The risk was highest following focal cerebral injury and increased with the severity of the TBI (Table 2). The probability of an epilepsy diagnosis was not significantly higher following mild injury or fracture than in the non-TBI group (Figure 2B).

Figure 2. Kaplan-Meier Curves Illustrating the Risk of Epilepsy Following a First Seizure.

(A) Divided by occurrence of traumatic brain injury (TBI) before the seizure. TBI (red) and no TBI (blue). (B) Divided by the severity of the TBI before the seizure. No prior TBI (blue), mild injury (green), fracture (purple), extracerebral (yellow), diffuse cerebral (black), and focal cerebral (orange).

Table 2.

10-Year Kaplan-Meier Estimated Risk of Epilepsy Following TBI and a Single Seizure

	Overall	<2 y following TBI	>2 y following TBI	Significance
	% (95% CI)	% (95% CI)	% (95% CI)	p Value
Non-TBI	33.4 (30.3–36.5)	31.2 (24.5–37.9)	33.2 (29.7–36.7)	Not significant
Any injury	41.1 (38.6–43.7)	48.7 (44.8–52.6)	35.8 (32.3–39.3)	<0.001
Mild	33.9 (30.8–37.0)	42.6 (37.5–47.7)	29.1 (25.2–33.0)	<0.001
Fracture	40.5 (28.4–52.7)	47.8 (28.2–67.4)	37.0 (21.9–52.1)	Not significant
Extracerebral	51.2 (42.2–60.2)	51.2 (40.2–62.2)	50.5 (33.1–67.9)	Not significant
Diffuse cerebral	54.3 (46.5–62.1)	59.3 (48.9–69.7)	50.7 (38.4–63.1)	Not significant
Focal cerebral	62.3 (53.7–70.9)	64.8 (53.6–76.0)	60.1 (47.4–72.8)	Not significant

Abbreviation: TBI = traumatic brain injury.

Overall, and stratified for the time of the TBI to first seizure, including the significance level for the time stratification. Potential significant differences were measured between <2 y following TBI and >2 y following TBI.

Time From Index Date

Since other forms of acquired/remote symptomatic epilepsy typically develop in the first year or 2 after the brain insult,^2,7,11 we performed a stratified analysis depending on whether the first PTS occurred <2 years or >2 years following the TBI. For individuals with a first PTS <2 years following the TBI (n = 1,140), the risk of an epilepsy diagnosis was 48.7% (95% CI 44.8–52.6). This risk was significantly higher (p < 0.001) than in individuals with their first PTS >2 years following the TBI (n = 3,099)—35.8% (95% CI 32.3–39.3). When stratifying for trauma severity, we found that the risk of epilepsy was significantly higher following mild injury if the seizure occurred <2 years following the trauma (Table 2).

Hazard of Epilepsy

Since the studied groups were not age-matched and sex-matched, and of other confounding factors, we performed Cox proportional hazards modelling. In addition to an analysis of the entire time span, we also performed time-stratified analyses because the Kaplan-Meier curves indicated varying risks of epilepsy throughout the study period. The time strata used were years 0–1, years 1–2, and years 2–10. These models agreed with the main results from the Kaplan-Meier analyses; there was no significantly elevated HR for epilepsy following mild injury or fracture for any of the time strata with adjustments for age, sex, and comorbidities, whereas the risk of epilepsy was increased for the more severe TBI forms (Table 3).

Table 3.

Cox Proportional Hazard Model of Epilepsy Compared With the Non-TBI Group, Unadjusted and With Adjustments for Age, Sex, and Time-Dependent Adjustments for CNS Comorbidities

Entire follow-up	Unadjusted HR (95% CI)	Significance p value	Adjusted HR (95% CI)	Significance p value
Non-TBI	Reference	Reference	Reference	Reference
Mild	1.0 (0.8–1.1)	Not significant	1.0 (0.8–1.1)	Not significant
Fracture	1.2 (0.9–1.7)	Not significant	1.2 (0.9–1.7)	Not significant
Extracerebral	1.5 (1.2–1.8)	<0.001	1.4 (1.1–1.7)	0.005
Diffuse cerebral	1.9 (1.5–2.3)	<0.001	1.9 (1.5–2.3)	<0.001
Focal cerebral	2.2 (1.8–2.8)	<0.001	2.1 (1.7–2.7)	<0.001
Year 1
Mild	0.8 (0.7–1.0)	Not significant	0.8 (0.7–1.0)	0.018
Fracture	1.2 (0.8–1.8)	Not significant	1.1 (0.7–1.8)	Not significant
Extracerebral	1.3 (1.0–1.7)	Not significant	1.2 (0.9–1.6)	Not significant
Diffuse cerebral	1.9 (1.4–2.4)	<0.001	1.8 (1.4–2.3)	<0.001
Focal cerebral	2.0 (1.6–2.7)	<0.001	1.9 (1.5–2.5)	<0.001
Year 2
Mild	1.3 (0.9–1.9)	Not significant	1.4 (0.9–2.0)	Not significant
Fracture	1.8 (0.8–4.1)	Not significant	1.8 (0.7–4.1)	Not significant
Extracerebral	2.0 (1.2–3.5)	0.012	1.7 (1.0–3.0)	Not significant
Diffuse cerebral	1.9 (1.0–3.4)	0.035	1.9 (1.0–3.4)	0.034
Focal cerebral	2.3 (1.2–4.4)	0.008	2.2 (1.1–4.1)	0.019
Years 3–10
Mild	1.2 (0.9–1.7)	Not significant	1.3 (1.0–1.8)	Not significant
Fracture	1.1 (0.5–2.5)	Not significant	1.1 (0.5–2.6)	Not significant
Extracerebral	2.1 (1.3–3.2)	0.002	1.7 (1.1–2.8)	0.018
Diffuse cerebral	1.9 (1.2–3.1)	0.009	1.9 (1.2–3.2)	0.008
Focal cerebral	3.0 (1.8–4.8)	<0.001	2.8 (1.7–4.6)	<0.001

Abbreviation: TBI = traumatic brain injury.

In addition to time-stratified hazards for year 1, year 2, and years 3–10.

Discussion

We found a significantly increased risk of epilepsy among individuals with a first PTS, compared with the risk seen after first seizures without previous TBI. The highest risk was seen following focal cerebral injuries. In agreement with our hypothesis, the risk increased with the severity of the TBI, and was higher if the first seizure occurred within 2 years of the injury. We could not identify any significant risk difference between individuals following mild injury and skull fracture, compared with the non-TBI group. Estimating the recurrence risk is important, both for the initiation of ASM treatment and because the ILAE definition of epilepsy allows for an epilepsy diagnosis already after a first seizure.³ However, our injury classification did not identify a subgroup with a risk exceeding the ILAE threshold, highlighting a need for caution in cases of first PTS to avoid misdiagnosis of epilepsy. The risk of epilepsy in our cohort was somewhat lower compared with the majority of the literature,⁶ although we have included all types of TBI, unlike most previous studies, that focused on individuals at risk of PTE or with severe TBI.^4,11 The risk of PTE found in these studies was higher, even compared with the individuals with focal cerebral injuries in our study, which probably reflects our classification, which included all focal cerebral injuries and not only the most severe ones.

The risk of epilepsy was further increased if the first PTS occurred <2 years following the TBI. This closely resembles the situation in poststroke epilepsy⁷ and suggests that latency can be a prognostic indicator. Two years following the TBI is the time frame analyzed in the most recent studies on the recurrence risk of seizures after a first unprovoked PTS,^11,12 with a high risk of seizure recurrence, motivating an epilepsy diagnosis already after a first unprovoked seizure.³ In our study, the risk of seizure recurrence tended to be lower across all trauma severity levels if the first PTS occurred >2 years following the trauma. This further highlights that the risk of seizure recurrence even after the most severe TBI may not motivate an epilepsy diagnosis, if the first unprovoked PTS occurs >2 years following the trauma. Of interest, our time-segmented analysis also showed that the risk was higher following mild TBI if the first PTS occurred early. Speculatively, this may reflect that some individuals diagnosed with mild TBI do sustain a permanent brain injury because of their TBI. Alternatively, the result may be an effect of residual confounding factors that increase the risk of both epilepsy and trauma, such as substance abuse.^13,14

The strengths of the study include the size of the cohort, which encompasses all individuals hospitalized with a TBI, using data from a high-accuracy register,¹⁵ in which the diagnostic accuracy of an epilepsy diagnosis has been found to be over 90%.⁹ Epilepsy in general is diagnosed by neurologists, which we believe is a reason for the high accuracy of the diagnosis. The large dataset enabled us to perform stratified analyses to identify risks in smaller cohorts. The coverage of the register also limits loss to follow-up because the register is linked to the individual identification number, allowing for a follow-up time of up to 18 years.

Our register-based approach is further validated by some key findings. First, the risk of epilepsy in the non-TBI group was similar to the overall risk of epilepsy after a first unprovoked seizure.¹⁶ In addition, the overall risk in the TBI group was in line with the results of an American-conducted study with similar study design.⁵ Finally, 92% of the individuals diagnosed with epilepsy were prescribed ASM in close connection with the diagnosis, indicating high accuracy of the epilepsy diagnosis.¹⁰

There are also limitations because our method only included individuals diagnosed with a first PTS and is therefore unable to capture individuals who are diagnosed with epilepsy without a prior seizure diagnosis, for instance, if multiple seizures occur before presenting to health services. In individuals with more severe TBI, it is also possible that clinicians diagnosed epilepsy after a single seizure, potentially leading to a falsely lower risk of epilepsy in that group in our cohort, although in our experience, this was not a widespread practice in Sweden during the study period.

Furthermore, we have not performed any validation analyses of the ICD-10 codes for TBI. However, we believe that the effects of a potential miscoding between the trauma severities may lead to a nondifferential misclassification, which may at worst slightly dilute results.

In our review of medical records, we found that almost 30% of first PTS were acute symptomatic, which is higher than previously reported.¹⁷ Many acute symptomatic seizures were attributed to substance abuse, warranting particular clinical vigilance in this patient group.^{14 8} The high proportion of acute symptomatic seizures will have made our epilepsy estimates more conservative because acute symptomatic seizures have a lower risk of seizure recurrence than unprovoked seizures.⁵ It should however not affect our main analyses of differences between trauma severities more than marginally. To further validate our main findings, we have also performed a Cox proportional hazard model to adjust for factors that potentially influence the risk of epilepsy.

It is important that the overall risk of epilepsy following mild injury or fracture was not increased compared with the non-TBI group. The skull fracture group included 111 individuals with ensuing uncertainty in PTE risk estimates. However, since the overall risk of epilepsy following TBI is almost identical following mild injury and fracture,² and none of these groups imply a direct cerebral injury, we find it reasonable that the risk following one PTS would be the same in these groups. We therefore conclude that neither mild injury nor fracture is a certain indicator of an increased risk of epilepsy after a first PTS. More prognostic indicators are needed. For instance, it would be interesting to analyze the importance of injury location for epilepsy risk, an analysis that is not possible using our register data. Another interesting group would be those with status epilepticus as their first seizure, where more data are needed to guide management. Remote nonprogressive causes had a 5-year recurrence risk of 34% according to a recent study,¹⁸ but our results suggest that this may vary between patients with different TBI severity. In addition, other ways of using administrative data to analyze more detailed aspects may be an interesting approach for further research, such as natural language processing.¹⁹ Our results suggest that indicators enabling an epilepsy diagnosis already after a first seizure should be sought in the more severe TBI groups, particularly if the first PTS occurs soon after the TBI, and that individuals with milder TBI and a first PTS should be counselled largely as those with a first seizure without a previous TBI regarding the risk of epilepsy. Establishment of predictive models incorporating clinical data on timing and TBI characteristics, similar to the SeLECT model for poststroke epilepsy,²⁰ could also be worth exploring—although the heterogeneity of TBI may make the challenge more difficult than in stroke.

In conclusion, we found that the risk of epilepsy following a first PTS increases with the severity of the TBI and is higher if there is less than 2 years between the TBI and the PTS. More research on prognostic indicators, preferably objective biomarkers, are needed.

TAKE-HOME POINTS

• → Severe traumatic brain injury (TBI) increases the risk of epilepsy substantially.

• → After a single posttraumatic seizure (PTS) following mild TBI, the risk of epilepsy does not seem to be increased.

• → The risk of epilepsy is further increased if the first PTS occurs <2 years following the trauma.

Appendix. Authors

Name	Location	Contribution
Markus Karlander, MD	Institute of Neuroscience and Physiology, Department of Clinical Neuroscience, Sahlgrenska Academy, Gothenburg University; Clinic of Neurology, Rehabilitation and Specialized Care at Home, Södra Älvsborg Hospital; Department of Research, Education and Innovation, Region Västra Götaland, Södra Älvsborg Hospital, Borås, Sweden	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Samuel Håkansson, PhD	Institute of Neuroscience and Physiology, Department of Clinical Neuroscience, Sahlgrenska Academy, Gothenburg University; Department of Neurology, Sahlgrenska University Hospital; Wallenberg Center of Molecular and Translational Medicine, Gothenburg University, Sweden	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Johan Ljungqvist, PhD	Institute of Neuroscience and Physiology, Department of Clinical Neuroscience, Sahlgrenska Academy, Gothenburg University; Department of Neurosurgery, Sahlgrenska University Hospital, Gothenburg, Sweden	Drafting/revision of the manuscript for content, including medical writing for content
Ann Sörbo, PhD	Institute of Neuroscience and Physiology, Department of Clinical Neuroscience, Sahlgrenska Academy, Gothenburg University; Clinic of Neurology, Rehabilitation and Specialized Care at Home, Södra Älvsborg Hospital; Department of Research, Education and Innovation, Region Västra Götaland, Södra Älvsborg Hospital, Borås, Sweden	Drafting/revision of the manuscript for content, including medical writing for content
Johan Zelano	Institute of Neuroscience and Physiology, Department of Clinical Neuroscience, Sahlgrenska Academy, Gothenburg University; Department of Neurology, Sahlgrenska University Hospital; Wallenberg Center of Molecular and Translational Medicine, Gothenburg University, Sweden	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data

Study Funding

The study was funded by the Swedish Society of Medical Research: S18-0040; Linnea and Josef Carlsson foundation: 90_20180321_048; The Promobilia foundation: 18012; The Swedish state through the ALF agreement (ALFGBG-965029); and Södra Älvsborg Hospital—Department of Research, Education and Innovation: sas-981863.

Disclosure

J. Zelano: reports speaker honoraria from UCB and Eisai, writers honoraria/royalty from Neurologi i Sverige/Liber AB, Studentlitteratur AB, and as an employee of Sahlgrenska University Hospital (no personal compensation) being investigator in clinical trials sponsored by UCB, Bial, SK life science, and GW Pharma. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/cp.