Risk of Epilepsy and Factors Associated With Time to Seizure Remission in Anti-LGI1 Encephalitis

Abstract

Background and Objectives

Autoimmune encephalitis (AIE) with anti-leucine-rich glioma-inactivated 1 (LGI1) antibodies typically manifests with subacute cognitive deficits, seizures, and psychiatric symptoms, mostly in older adults. Immunotherapy (IT) leads to the cessation of seizures in most patients, yet some develop AIE-associated epilepsy (AEAE) and persistent cognitive deficits. The aim of this large multicentric retrospective observational cohort study was to assess long-term outcomes of patients with anti-LGI1 encephalitis regarding seizures and AEAE and to identify associated factors.

Methods

We included patients with anti-LGI1 encephalitis from 3 national referral centers/consortia meeting the following inclusion criteria: (I) definite LGI1 limbic encephalitis (Graus criteria); (II) occurrence of seizures; and (III) follow-up period ≥24 months. We aimed to (1) determine the risk of seizure recurrence (ROSR) on remission, (2) investigate clinical and paraclinical biomarkers for an effect on time to seizure remission using Cox proportional hazard modeling (n = 188), and (3) assess the risk of AEAE and determine associated factors (n = 236).

Results

AEAE was observed in 5.9% (16/271) of the full cohort. Both AEAE (16/16 vs 129/215, p = 0.001) and longer time to seizure remission (OR 1.36 per year, p = 0.025) were associated with persistent cognitive impairment. Patients with pilomotor seizures had a lower rate of seizure remission (hazard ratio [HR] 0.58, 95% CI 0.55–0.60, p < 0.001) while patients under IT administration had a higher rate of seizure remission over time (HR 12.4, 95% CI 9.67–16.0, p < 0.001). In addition, patients receiving second-line IT tended to achieve earlier seizure remission (log-rank test, p = 0.019). The ROSR at 12, 60, and 120 months on seizure remission was 9% (95% CI 4.5%–13%), 20% (95% CI 11%–28%), and 53% (95% CI 14%–74%), respectively.

Discussion

In conclusion, our results demonstrate that AEAE in anti-LGI1 encephalitis is rare and suggest that the diagnosis of epilepsy is inappropriate in patients reaching seizure remission because of a relatively low ROSR. Accordingly, on seizure remission, the diagnosis of acute symptomatic seizures would be appropriate. Moreover, we validate and quantify the importance of IT for seizure remission and identify biomarkers associated with lower rates of seizure remission. Late remission of seizures and AEAE were associated with persistent cognitive impairment.

Introduction

Patients with autoimmune encephalitis (AIE) characterized by antibodies targeting leucine-rich glioma-inactivated 1 (LGI1) typically present with a subacute onset of frequent epileptic seizures, cognitive impairment, and psychiatric symptoms, in many cases fulfilling the current diagnostic criteria for autoimmune limbic encephalitis.¹ In 28%–69% of the patients, faciobrachial dystonic seizures (FBDS), which are pathognomonic for this disease, occur early in the disease course.^2-4 Other common characteristics of the disease are pilomotor seizures or thermal sensations, paroxysmal dizziness spells, and hyponatremia.^5-7 Brain MRI shows unilateral or bilateral T2/fluid-attenuated inversion recovery (T2/FLAIR) hyperintensity of the temporomesial structures in 41%–74%.^7,8 As the disease progresses, hippocampal sclerosis (HS) or atrophy develops in 44%–77% of the patients.^8-10 The disease predominantly affects older patients and exhibits a strong association with the human leukocyte antigen (HLA) allele DRB1*07:01 (∼90% of carriers).^11-14 Seizures are typically resistant to antiseizure medications (ASMs) in the acute phase of the disease. However, multiple studies have demonstrated high response rates to immunotherapies (ITs), with most patients achieving sustained seizure freedom.^15,16 Although seizures respond well to IT, recovery of cognitive function seems to be less favorable.^8,17-19 Indeed, some patients not only have incomplete cognitive recovery but also continue to have epileptic seizures years after the acute phase, fulfilling criteria for epilepsy or rather AIE-associated epilepsy (AEAE).^20-22 Most studies reported a low prevalence of AEAE after anti-LGI1 encephalitis (around 2%–20%).^4,11,23,24 However, a recently published study reported a prevalence of up to 72%.¹⁰ Unfortunately, in these studies, no consistent diagnostic criteria for AEAE were applied. The objectives of our study were to (1) estimate the risk of seizure recurrence (ROSR) upon seizure remission to determine whether the diagnosis of acute symptomatic seizures is appropriate once seizures go into remission, (2) identify clinical and paraclinical biomarkers associated with time to seizure remission, and (3) assess the prevalence of patients with AEAE and factors associated with AEAE.

Methods

Study Design and Participants

This retrospective cohort included patients from 3 national networks/reference centers: (I) German Network for Research on Autoimmune Encephalitis (GENERATE; en.generate-net.de; German cohort), (II) French Rare Disease Reference Centre for Paraneoplastic Neurological Syndromes and Autoimmune Encephalitis (French cohort), and (III) Dutch National Referral Site for Patients with Suspected Autoimmune Encephalitis (Dutch cohort). The study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.²⁵

The primary cohort of patients analyzed included 87 patients from the French cohort and 101 from the German cohort. All patients met the following criteria: (1) definite LGI1 limbic encephalitis according to Graus criteria,¹ (2) occurrence of epileptic seizures, and (3) a follow-up period ≥24 months. To increase the sample size of the AEAE factor analysis, we further recruited an enriched Dutch cohort (extension cohort) adhering to these inclusion criteria, consisting of all patients from this cohort who experienced seizures up to their latest follow-up (n = 12), along with three times the number of age-matched and sex-matched controls (total: n = 48, Figure 1). To prevent a potential selection bias resulting from this enrichment for patients with persisting seizures, the Dutch cohort was excluded from the remaining analyses.

Figure 1. Composition of the Study Population.

Flowchart illustrating the inclusion and composition of patients included in the study.

Antibody testing was conducted using commercially available cell-based assays and, in most cases, confirmed using standard tissue-based immunofluorescence and/or immunohistochemistry procedures (eTable 1).^3,26,27 Positive results were regarded as specific, irrespective of antibody titer.

For all patients, we acquired comprehensive data including sex, age at disease onset, initial clinical symptom, time of first and last seizure, the duration of seizure freedom with and without ASMs, and the modified Rankin Scale (mRS) score at first and last follow-up. All periods were assessed at a resolution of “months” (month 0: 0–15 days, month 1: 16–45 days, etc). In addition, we assessed tumor diagnoses and the presence of cognitive deficits at the last follow-up visit. The assessment of cognitive deficits was based on the tests performed in the individual centers and the clinical impression of the practitioner. Regarding seizures, we collected the following information: occurrence of FBDS, pilomotor seizures, and bilateral tonic-clonic seizures (Table 1).

Table 1.

Characteristics of Cohorts

	Primary cohort		Individual
Characteristic	N	N = 188a (%)	GER, N = 101a (%)	FR, N = 87a (%)	NL, N = 48a (%)
Female	188	73 (39)	43 (43)	30 (34)	20 (42)
Age at disease onset (y)	188	64 (56–72)	64 (53–71)	65 (57–73)	64 (55–69)
Follow-up duration (mo)	188	44 (31–69)	48 (32–77)	41 (30–58)	62 (43–88)
Time until diagnosis (mo)	188	4 (1–8)	4 (1–9)	4 (2–8)	5 (2–8.5)
Latency to IT after first symptoms (mo)	187	4 (2–8)	4 (1–9)	4.5 (2–8)	5 (2–8)
Seizures 12 mo after first symptomsb	188	83 (44)	52 (51)	31 (36)	24 (50)
Seizures 12 mo after IT initiationb	187	42 (22)	30 (30)	12 (14)	16 (35)
Seizure remission at last follow-up	188	169 (90)	89 (88)	80 (92)	36 (75)
Seizure remission at last follow-up without ASMs	187	57 (30)	36 (36)	21 (24)	25 (52)
Clinical relapse	188	49 (26)	27 (27)	22 (25)	16 (33)
Seizure recurrence	188	29 (15)	9 (8.9)	20 (23)	4 (8.3)
First clinical symptom	188
Faciobrachial dystonic seizures		31 (16)	14 (14)	17 (20)	8 (17)
Seizures		71 (38)	45 (45)	26 (30)	22 (46)
Cognitive impairment		62 (33)	28 (28)	34 (39)	15 (31)
Psychiatric symptoms		20 (11)	10 (9.9)	10 (11)	1 (2.1)
Other		4 (2.1)	4 (4)	0 (0)	2 (4.2)
Faciobrachial dystonic seizures	188	83 (44)	35 (35)	48 (55)	20 (42)
Pilomotor seizures	188	32 (17)	23 (23)	9 (10)	16 (33)
Bilateral tonic-clonic seizures	188	97 (52)	52 (51)	45 (52)	21 (44)
Persistent cognitive impairment	188	116 (62)	61 (60)	55 (63)	33 (69)
First mRS score	187	3 (2–3)	3 (2–3)	2 (1–3)	3 (2–3)
Last mRS score	187	1 (1–2)	1 (1–2)	1 (0–2)	2 (1–2)
Anti-LGI1 antibodies in CSF	178	130 (73)	58 (62)	72 (86)	28 (68)
Pleocytosis in CSF	179	20 (11)	8 (8.6)	12 (14)	5 (12)
Oligoclonal bands in CSF	139	15 (11)	8 (8.9)	7 (14)	3 (21)
Hyponatremia	188	78 (41)	30 (30)	48 (55)	22 (46)
Tumor diagnosis	188	17 (9)	8 (7.9)	9 (10)	7 (15)
Initial MRI suggestive of limbic encephalitisc	181
Unilateral		69 (38)	47 (48)	22 (26)	11 (23)
Bilateral		73 (40)	29 (30)	44 (52)	25 (52)
Hippocampal sclerosis in latest follow-up MRI	155
Unilateral		43 (28)	24 (26)	19 (30)	5 (24)
Bilateral		36 (23)	15 (16)	21 (33)	3 (14)
Epileptiform discharges	182
Unilateral		62 (34)	32 (32)	30 (37)	15 (35)
Bilateral		17 (9.3)	16 (16)	1 (1.2)	7 (16)
ASM during the course of the disease	188	177 (94)	93 (92)	84 (97)	41 (85)
Immunotherapy	188	187 (99)	101 (100)	86 (99)	46 (96)
Corticosteroid therapy	188	167 (89)	98 (97)	69 (79)	44 (92)
Intravenous immunoglobulin	188	115 (61)	37 (37)	78 (90)	34 (71)
Plasma exchange/immunoadsorption	188	49 (26)	47 (47)	2 (2.3)	2 (4.2)
Second-line IT	188	95 (51)	38 (38)	57 (66)	12 (25)
Rituximab	188	85 (45)	36 (36)	49 (56)	12 (25)
Cyclophosphamide	188	47 (25)	5 (5)	42 (48)	0 (0)
Steroid-sparing long-term IT	188	43 (23)	25 (25)	18 (21)	34 (71)

Abbreviations: ASM = antiseizure medication; FR = France; GER = Germany; IT = immunotherapy; mRS = modified Rankin Scale; NL = Netherlands.

^an (%); median (interquartile range).

^bOnly considering the first period with seizures, i.e., without accounting for seizure recurrences.

^cHyperintense signal of the temporomesial structures in T2/fluid-attenuated inversion recovery sequences.

We defined remission of seizures as achieving seizure freedom for ≥6 months during the follow-up period and AEAE as the ongoing occurrence of seizures until the last follow-up for ≥24 months after initiating IT, with a negative result for LGI1 antibodies in serum during the most recent follow-up test.²¹ A period of seizure freedom was labeled as such only if remission of seizures was reached. Similarly, recurrence of seizures required achieving seizure remission beforehand. For all patients meeting the AEAE criteria, the prescribed ASMs were assessed regarding drug-resistant epilepsy.²⁸

Clinical relapses were defined as an mRS score increase of ≥1 point during the follow-up and/or IT re-administration after an initial improvement. We assessed the application of various ITs throughout the disease course and the timing of the first IT administration. Corticosteroids, IV immunoglobulin (IVIG), and plasma exchange/immunoadsorption (PLEX/IA) were classified as first-line ITs, while cyclophosphamide and/or rituximab were classified as second-line ITs.⁴ Azathioprine, mycophenolate, and methotrexate were considered long-term steroid-sparing ITs.

The following radiologic and electroencephalographic findings were recorded: absent, unilateral or bilateral (1) T2/FLAIR hyperintensities of the temporomesial structures in the initial MRI, (2) HS in the most recent MRI scan and (3) epileptiform discharges. Results of CSF analysis (white cell count, oligoclonal bands) and the presence of hyponatremia were also noted. Follow-up serum antibody testing data was only obtained for the subset of patients who had recurrent seizures for ≥24 months after initiating IT. In addition, we recorded whether the last MRI of these patients showed signs of acute autoimmune limbic encephalitis.

In a subset of 55 German patients, neurofilament light chain (NfL) levels were determined from available CSF and serum (detailed methods in eFigure 1). Serum NfL values from 598 healthy participants served as controls, comprising individuals from the Memory and Morbidity in Augsburg Elderly study²⁹ and healthy blood donors recruited at Kiel University. In addition, CSF from 68 healthy control patients was analyzed.

Statistical Analysis and Graphics

Statistical analysis was performed with R 4.3.3.³⁰ Categorical data was summarized in absolute and relative frequencies. Continuous data was summarized using “median (interquartile range [IQR]).” Wilcoxon rank-sum tests were used to compare unpaired numeric data. Frequencies of categorical variables were compared using Pearson's χ² test or Fisher's exact test as appropriate. For comparing NfL measurements, analysis of covariance adjusting for age was used after log₂ transformation of the NfL concentrations to achieve an approximately normal distribution.

Kaplan-Meier estimators with accompanying log-rank tests were used to identify biomarkers associated with seizure remission. A multivariable Cox proportional hazard regression model was applied to adjust for covariates (included covariates in Table 2) using survival 3.7.0.³¹ To adjust for possible cohort effects, cohort (German or French) was used as a clustering parameter of this Cox model. An additional Cox regression treating cohort as a fixed effect was conducted as a sensitivity analysis (eTable 2). Because of 24 missing out of 2,068 total values (1.2%) in the data for the Cox regression, mice 3.17.0 was used to impute these values for this regression.³² The results based on these imputations were pooled using the Rubin rule. For other analyses, patients with missing data were excluded. ROSR was estimated with ggsurvfit 1.1.0 (risk = 1–[Kaplan-Meier estimate]).³³

Table 2.

Cox Regression for Seizure Remission

Independent variable	HRa	Statistic	95% CI	p Value
Female	1.24	3.6	1.10–1.40	<0.001
Age at disease onset (y)b	1.01	2.1	1.00–1.02	0.039
Immunotherapy administrationc	12.4	20	9.67–16.0	<0.001
Faciobrachial dystonic seizures	0.77	−1.9	0.58–1.01	0.062
Pilomotor seizures	0.58	−26	0.55–0.60	<0.001
First mRS score ≥3	0.78	−3.6	0.68–0.89	<0.001
Anti-LGI1 antibodies in CSF	1.09	0.47	0.76–1.56	0.641
Hyponatremia	1.06	0.96	0.94–1.21	0.338
Initial MRI suggestive of limbic encephalitisd
Unilateral	1.03	0.18	0.73–1.46	0.861
Bilateral	0.86	−1.7	0.73–1.03	0.097
Epileptiform discharges
Unilateral	1.13	0.81	0.84–1.52	0.420
Bilateral	0.58	−5.0	0.47–0.72	<0.001

Abbreviation: mRS = modified Rankin Scale.

Number of events = 180. Results of 10,000 imputation models pooled for 24 missing out of 2,068 total values. Cohort was used as a clustering variable.

^aHazard ratio; lower HR corresponds to a lower rate of seizure remission.

^bNumeric.

^cTime-varying covariate.

^dHyperintense signal of the temporomesial structures in T2/fluid-attenuated inversion recovery sequences.

Adjusting for cohort, age, and sex, we used logistic regression to model the influence of time to seizure remission on either HS or persistent cognitive impairment.

Statistical tests were 2-tailed and considered statistically significant at p ≤ 0.05. All p values were rounded to the third decimal place (lower values shortened to “p < 0.001”) and reported without adjustment for multiple testing.

The code for producing the material, except for Figure 1, and statistical test results is available at github.com/freyberg/2025-lgi1-analysis. For data wrangling, plotting, and summary table generation, tidyverse 2.0.0, ggsurvfit 1.1.0, and gtsummary 2.0.4 functions were used.^33-35

Standard Protocol Approvals, Registrations, and Patient Consents

Informed written consent was obtained from each patient with approval of the local ethics committees of participating centers (Ethics Committee of the University of Lübeck, Germany [reference number: 13–162]; Institutional Review Board of the Erasmus Medical Center (MC); and Institutional Review Board of the Hospices Civils de Lyon [NCT04106765]), and all procedures were conducted in accordance with the Declaration of Helsinki.

Data Availability

Data supporting these findings are available from the corresponding authors on reasonable request.

Results

Patient Characteristics

The primary cohort comprised 188 patients who met the inclusion criteria (Figure 1). The median follow-up period was 44 months (IQR 31–69) (Table 1). There was a slight male predominance, and the median age at onset was 64 years (IQR 56–72). The most frequent symptoms at onset were seizures (including FBDS) and cognitive deficits. Patients with FBDS frequently experienced seizures with other semiologies (20/35 patients [57%], data only available for the German cohort).

Brain MRI data at disease onset revealed unilateral or bilateral temporomesial T2/FLAIR hyperintensity in 142 of 181 patients (78%; Table 1). In follow-up MRI scans, 79 of 155 patients (51%) showed unilateral or bilateral HS. Unilateral or bilateral epileptiform discharges were detected in 79 of 182 patients (43%). Of the 100 German patients undergoing EEG, epileptiform discharges were detected more frequently in patients who underwent long-term EEG compared with those who underwent routine EEG only (long-term: 27/40 [68%] vs routine: 21/60 [35%], χ² = 10.156, df = 1, p = 0.001). Antibodies against LGI1 were detected in serum in all patients and in CSF in 130 of 178 patients (73%) (eTable 1).

NfL levels were assessed in a subgroup of 55 German patients (110 serum and 35 CSF samples, median: 2 serum samples per patient, IQR 1–3). Although we confirmed previous observations of increased CSF NfL levels³⁶ and further showed elevated serum NfL levels (serum: p < 0.001, CSF: p = 0.001; eFigure 1, A and B) early in the disease course, late NfL samples were no longer different from those of controls (serum: p = 0.193, CSF: p = 0.35; eFigure 1, C and D). We did not observe associations of CSF or serum NfL levels with seizure remission or outcome (not shown).

Immunosuppressive Treatment

The median duration from onset of seizures to initiation of IT was 3 months (IQR 1–7). Second-line IT was administered more frequently in the French cohort than in the German cohort (Table 1), indicating varying treatment preferences (57/87 [66%] vs 38/101 [38%], χ² = 14.548, df = 1, p < 0.001).

We also investigated the influence of different first-line ITs as well as the use of second-line IT and steroid-sparing long-term ITs on seizure remission (≥6 months of seizure freedom). No differences were observed among the first-line therapies (log-rank tests, corticosteroids: p = 0.246, IVIG: p = 0.492, PLEX/IA: p = 0.128). However, patients who had received second-line IT had a higher rate of seizure remission (Figure 2, eFigure 2A). Owing to higher second-line IT usage in the French cohort, we further investigated the effect of second-line IT on seizure remission for each cohort separately but found no cohort-specific differences (eFigure 2, B–D). Rituximab showed a similar effect to that of any second-line IT in the primary cohort, while the same could not be demonstrated for cyclophosphamide (eFigure 2, E–F) or steroid-sparing long-term IT (log-rank test, p = 0.874).

Figure 2. Kaplan-Meier Plot for Seizure Remission Comparing Patients Who Did or Did Not Receive Second-Line Immunotherapy (IT).

Primary Cohort. p Value based on corresponding log-rank test. Risk table below plot. For better readability, the x-axis is cut at 76 months. A plot showing all events is available in eFigure 2A.

Course of Seizures

In 142 of 188 patients (76%), seizures manifested within the initial 2 weeks of the disease. Seizure freedom often occurred rapidly upon IT initiation, with 180 of 188 patients (96%) achieving it at least once (Figure 3A). Seizure freedom was attained after a median time of 4 months (IQR 0–10) upon IT initiation, with 48 of 187 patients (26%) becoming seizure-free within the same month and 109 of 187 patients (58%) within 6 months of IT initiation (Figure 3B-C and eFigure 3). Still, 42 of 187 (22%) and 21 of 187 (11%) patients experienced seizures 12 and 24 months after the initiation of IT, respectively. The time to seizure remission after IT initiation was not associated with the latency to IT initiation, i.e., the duration of the symptomatic period without IT (log-rank test, p = 0.757). Across both cohorts, a substantial number of patients continued to take ASMs until their last follow-up visit despite having achieved seizure remission (53/89 [60%] and 58/79 [73%] patients in the German and French cohort, respectively).

Figure 3. Time to Seizure Freedom and Initiation of Immunotherapy (IT).

(A) Bar chart illustrating the occurrence of seizures in each patient of the German and French cohort throughout the course of the disease, filtered for 187 of 188 patients (99%) who received immunotherapy (IT) at some point. The vertical line at t = 0 months indicates the time of IT initiation. Time with first symptoms besides seizures is indicated in brown, seizures in orange, seizure freedom in light blue and seizure freedom without ASMs in light green, sorted by time with seizures after starting IT. For periods with a duration of (rounded) 0 months, 0.5 months were added for visualization. Seizure-free intervals <6 months were not assessed. ASM = antiseizure medication. (B) Histogram of the time of (first) seizure freedom relative to the time of first IT administration (t = 0 months). Note that some patients are not shown for better readability (n = 2 with seizure freedom 5 months before and n = 14 more than 25 months after IT initiation). (C) Jitter plot of the time of first IT administration and the time to seizure freedom (relative to first symptoms; jitter of 0.2 in all directions to reduce overlap, opacity: 0.5). For better readability, some patients are not shown (n = 5 with seizure freedom after more than 53 months and n = 4 with IT initiation more than 20 months after seizures started). eFigure 3 includes corresponding plots for (B) and (C) showing all patients.

Recurrence of seizures was observed in 29 of 180 patients (16%) with seizure remission. Among these patients, seizures recurred after a median of 18 months (IQR 11–28) of seizure freedom. Patients reaching seizure remission had a ROSR of 9% (95% CI 4.5%–13%), 16% (95% CI 9.3%–21%), 20% (95% CI 11%–28%) and 53% (95% CI 14%–74%) at 12, 36, 60 and 120 months thereafter, respectively (Figure 4). Although associated with a high degree of uncertainty at later follow-up time points, the point estimate of the risk, therefore, remained below 60% for 10 years. Until 95 months, even the 95% CI of the risk estimate remained below 60% (ROSR at 95 months: 29%, 95% CI 13%–43% vs ROSR at 96 months: 41%, 95% CI 11%–61%). Furthermore, it should be noted that patients experiencing a seizure recurrence had longer follow-up durations overall (recurrence: 54 months, IQR 41–84, vs no recurrence: 42 months, IQR 29–66, W = 3,008.5, p = 0.009) and also after seizure remission (40 months, IQR 28–60, vs 30 months, IQR 21–48, W = 2,828.5, p = 0.013), thereby potentially skewing calculations toward higher estimates at later time points. In 10 of 29 patients (34%), the treating physicians did not consider the recurrence of seizures to be related to an anti-LGI1 encephalitis relapse because neither a one-point mRS increase nor an IT re-administration was documented. In the remaining cases, it was difficult to retrospectively determine whether the recurrence was (assumed to be) associated with an AIE relapse.

Figure 4. Risk of Seizure Recurrence .

The risk of seizure recurrence (ROSR) (and its 95% CI shaded area) is calculated by subtracting the Kaplan-Meier estimate (KM) for the event of seizure recurrence (and its CI) from 1 (ROSR = 1–KM), filtered for 180 of 188 patients (96%) who reached seizure remission (≥6 months of seizure freedom) once. Risk table below plot. A gray dashed line marks a ROSR of 60%. Note that the point estimate of the ROSR does not surpass this threshold at any time point. For better readability, the x-axis is cut at 100 months because no further patients reach the event of seizure recurrence after this time point (only censored data after 100 months). Accordingly, the ROSR remains 53% (95% CI 14%–74%) from 97 months until the last censoring at 154 months.

Remission of Seizures

We assessed various clinical and diagnostic factors in a Cox regression model for their association with the time to seizure remission (Table 2). We chose these factors because of (frequently) high availability at early time points of the disease course. The presence of pilomotor seizures, an initial mRS score ≥3, and the detection of bilateral epileptiform discharges were associated with lower probability of seizure remission. There was a mild “protective” effect of older age at disease onset and female sex. Notably, IT administration was a major factor favoring seizure remission.

As a sensitivity analysis, we evaluated an additional Cox regression model with the cohorts as a fixed effect instead of as a clustering variable although age at disease onset violated the proportional hazards assumption in this model (eTable 2). Age at disease onset, administration of IT, and the occurrence of pilomotor seizures were associated with time to seizure remission, with similar effect sizes across models. While sex and the bilateral detection of epileptiform EEG activity had no significant effect in this model, the occurrence of FBDS and the fixed effect of cohort (French cohort) influenced the rate of seizure remission compared with the original model.

In addition, we investigated associations between either persistent cognitive impairment or HS (summarizing unilateral or bilateral affection) and the time until seizure remission: The longer it took seizures to remit, the higher the chances of persistent cognitive impairment and HS in the last follow-up MRI scan (eTable 3). We further found that patients with persistent cognitive impairment at the last follow-up had had a lower probability of reaching seizure remission, while no difference emerged for patients with HS in the last available follow-up MRI (eFigure 4).

AIE-Associated Epilepsy

Finally, we investigated the frequency of and factors associated with AEAE in patients with anti-LGI1 encephalitis. Because only 7 of 188 patients (3.7%) in the primary cohort fulfilled the criteria for AEAE (with 5/101 [5%] and 2/87 [2.3%] patients from the German and French cohort, respectively), we sought to increase the number of patients with AEAE by adding an extension cohort to our analysis. In this extension cohort (Figure 1, Table 1, and eFigure 5), 83 patients met the inclusion criteria and 9 of 83 patients (11%) had AEAE.

Thus, 16 of 271 patients (5.9%) met the outlined AEAE criteria, while 250 of 271 patients (92%) could be classified as having had acute symptomatic seizures secondary to AIE. None of the 16 patients meeting the criteria for AEAE showed signs of active AIE in the most recent MRI (available for 12/16 [75%]). In addition, 10 of 16 patients (63%) with AEAE met the criteria for drug-resistant epilepsy,²⁸ with ≥2 ASMs at adequate doses having been administered (eTable 4). We were unable to categorize 5 of 271 patients (1.8%) into either category with certainty: these patients had seizures persisting until the last follow-up and for ≥24 months after starting IT, yet were still seropositive for LGI1 antibodies in the last available test (n = 4) or had unavailable data at the last follow-up (n = 1).

For the analysis of factors associated with AEAE, we excluded patients with unclear categorization (Table 3): Patients with AEAE were more likely to have persistent cognitive impairment and were younger at disease onset. Despite HS being present in 87 of 176 follow-up MRI scans (49%; available for 176/236 patients [75%]), only 8 of 87 patients (9.2%) met the criteria for AEAE. Second-line IT did not reach statistical significance for lower usage rates in patients with AEAE. Furthermore, there was no difference regarding the latency to IT initiation between patients with and without AEAE.

Table 3.

Comparison of Characteristics Between Patients With and Without AEAE

Characteristic	N	AEAE, N = 16a (%)	No AEAE, N = 215a (%)	p Value
Female	231	6 (38)	86 (40)	0.844b
Age at disease onset (y)	231	58 (48–63)	65 (57–71)	0.022c
Follow-up duration (mo)	231	65 (40–93)	47 (32–73)	0.033c
Latency to IT after first symptoms (mo)	228	4 (1–8.5)	4 (2–8)	0.745c
Clinical relapse	231	5 (31)	58 (27)	0.772d
Seizure recurrence	231	1 (6.3)	29 (13)	0.701d
Faciobrachial dystonic seizures	231	6 (38)	94 (44)	0.628b
Pilomotor seizures	231	6 (38)	40 (19)	0.098d
Bilateral tonic-clonic seizures	231	7 (44)	110 (51)	0.567b
First mRS score ≥3	230	12 (75)	123 (57)	0.170b
Persistent cognitive impairment	231	16 (100)	129 (60)	0.001b
Anti-LGI1 antibodies in CSF	214	9 (64)	146 (73)	0.538d
Pleocytosis in CSF	217	1 (6.7)	23 (11)	>0.999d
Oligocloncal bands in CSF	149	0 (0)	17 (12)	>0.999d
Hyponatremia	231	5 (31)	93 (43)	0.349b
Tumor diagnosis	231	0 (0)	22 (10)	0.376d
Initial MRI suggestive of limbic encephalitise	224	13 (87)	162 (78)	0.532d
Hippocampal sclerosis in latest follow-up MRI	171	8 (73)	75 (47)	0.097b
Epileptiform discharges	220	10 (63)	90 (44)	0.155b
Second-line IT	231	4 (25)	103 (48)	0.076b
Rituximab	231	4 (25)	93 (43)	0.153b
Cyclophosphamide	231	1 (6.3)	46 (21)	0.204d

Abbreviations: AEAE = autoimmune encephalitis–associated epilepsy; IT = immunotherapy; mRS = modified Rankin Scale.

^an (%); median (interquartile range).

^bPearson's χ² test.

^cWilcoxon rank-sum test.

^dFisher's exact test.

^eHyperintense signal of the temporomesial structures in T2/fluid-attenuated inversion recovery sequences.

Discussion

This large retrospective, multicentric, multinational study assessed remission of seizures, the occurrence of AEAE, long-term outcomes and their determining factors in patients with anti-LGI1 encephalitis. The major observations were as follows: (1) for patients achieving 6 months of seizure freedom, the diagnosis of acute symptomatic seizures secondary to AIE seems appropriate because of a low ROSR; (2) pilomotor seizures and bilateral epileptiform EEG activity were associated with lower rates of seizure remission over time; and (3) AEAE occurred in 16 of 271 patients (5.9%), with these patients more often having residual cognitive dysfunction.

The criteria of AEAE were met by 16 of 271 patients (5.9%). The number of patients included in our study markedly exceeds that of previous studies, whose cohorts ranged from 11 to 49 patients. These studies reported wide-ranging risk estimates for AEAE or chronic epilepsy (2%–72%), which underscores the persisting uncertainty about the risk of AEAE in patients with anti-LGI1 encephalitis. Some factors contributing to the wide range of risk estimates published thus far may lie in inconsistent definitions of epilepsy, different sample compositions and different follow-up periods across these studies.^{4,10,11,23,24,37} A potentially important factor leading to an overestimation of the risk of developing AEAE is the phenomenon of late seizure remission. While we find that most patients swiftly became seizure-free after the first IT administration, 22% and 11% of patients still had seizures 12 and 24 months later, respectively, hinting at the extended time frame after IT during which patients may still improve and become seizure-free. Accordingly, patients with late seizure remission and short follow-up may be misclassified as having epilepsy, based on the definitions used in those earlier case series, which reported higher incidence rates.^10,11,24 To address this issue, we applied a recently proposed practical definition of AEAE.²¹

Persistent cognitive deficits and young age at onset were linked to the presence of AEAE. Contrary to previous studies,^10,24 IT initiation was not delayed in patients with AEAE but rather effectively the same as for patients without AEAE. Previous case series yielded mixed results regarding an association between HS and the development of AEAE/chronic epilepsy in patients with anti-LGI1 encephalitis.^10,11,24 In general epilepsy cohorts, HS is a major risk factor for the development of drug-resistant epilepsy, with reported drug-resistant epilepsy rates ranging from 58% to 75%.^38,39 In our cohort, we found AEAE in only 8 of 87 patients (9.2%) with HS. The relatively low risk of AEAE despite HS in patients with anti-LGI1 encephalitis may indicate a pathophysiologic difference. Potentially related to this, previous studies observed focal atrophy in the CA3 and CA2/CA3 hippocampal subfield in patients with anti-LGI1 encephalitis.^18,40 This differs from the most common type obtained from epilepsy surgery patients, HS ILAE (International League Against Epilepsy) type 1, which is primarily characterized by CA1 and CA4 atrophy.⁴¹

Considering the pathophysiology of the disease, one may hypothesize that some proportion of patients with anti-LGI1 encephalitis may develop a chronic epilepsy responsive to ASMs, e.g., due to structural damage after inflammation has disappeared. This probably differs from other AEAE entities, such as Rasmussen's encephalitis or temporal lobe epilepsy associated with GAD65 antibodies, where chronic inflammatory factors might persist and most of the patients do not respond to classical ASMs.⁴² At the last follow-up, a large proportion of seizure-free patients continued to take ASMs across all 3 cohorts. It remains unclear whether the withdrawal of ASMs might be appropriate in many patients, e.g., those with seizure remission and an absence of anti-LGI1 antibodies. It is also conceivable that driving regulations influenced patients' decisions to continue ASMs.

Recurrence of seizures, regardless of its potential cause, was noted in approximately a sixth of patients after remission (patients being seizure-free for ≥6 months). The ROSR 5 years after seizure remission was lower in our primary cohort of patients with anti-LGI1 encephalitis (20%, 95% CI 11%–28%) than the ROSR 5 years after a first (33%, 95% CI 26%–40%) and after a second (73%, 95% CI 59%–87%) unprovoked seizure in the cohort given by Hauser et al.⁴³ Notably, the point estimate for the ROSR in patients with anti-LGI1 encephalitis remained below 60% at 120 months after remission but was associated with a high degree of uncertainty (53%, 95% CI 14%–74%). However, we demonstrated that the upper bound of the 95% CI did not surpass 60% until 95 months after seizure remission. In addition, in the context of patients with a seizure recurrence potentially skewing later estimates toward higher risk values because of longer follow-up durations, one may argue that this estimate is rather conservative and hypothesize that the true risk including uncertainty estimates (95% CI) may remain below 60% for 120 months, too. Taking into account the operational clinical definition of epilepsy by the ILAE, which requires a ROSR ≥60% after a first unprovoked seizure over the next 10 years,⁴⁴ the diagnosis of acute symptomatic seizures—rather than epilepsy—seems more appropriate for patients with anti-LGI1 encephalitis reaching seizure remission.

The criterion of ≥6 months of seizure freedom for our definition of seizure remission was guided by previous studies,^4,11 clinical experience and the attempt to increase the reliability of seizure absence in light of seizure underreporting.⁴⁵ However, it is possible that a shorter seizure-free interval is also sufficient.⁴⁶

Considering that a longer time until seizure remission was associated with a higher likelihood of experiencing cognitive impairment at the last follow-up and development of HS, it is a matter of considerable importance to identify clinical or diagnostic predictors of delayed remission of seizures. We identified pilomotor seizures as a clinical marker for a lower probability of seizure remission. One possible explanation for this finding may be that the inflammation and ensuing epileptogenic network spread beyond the temporomesial structures, as pilomotor seizures may originate from the parahippocampal gyrus or the median part of the insula.^47,48 We also identified bilateral epileptiform discharges and an initial mRS score ≥3 as biomarkers for reduced likelihood of seizure remission. These findings could similarly reflect a more extensively or more severely inflammation-affected network. Finally, younger age at disease onset and male sex were associated with later remission. Considering these findings, more vigilant monitoring and aggressive ITs should be considered in patients with these characteristics. The question of whether seizures themselves negatively affect cognition and promote the development of HS, or whether they serve as an effective biomarker for ongoing disease activity, remains unclear.

We provide further evidence for the pronounced efficacy of IT on seizure remission. Accordingly, late initiation of IT may postpone seizure remission, which, based on our data, increases the average risk of persistent cognitive deficits. No distinct advantage associated with any specific first-line IT was observed.¹⁶ Of interest, while prevalence rates of AEAE did not reach statistical significance for patients who received second-line IT, these patients had higher rates of seizure remission over time. This may provide tentative evidence for the efficacy of second-line IT in patients with anti-LGI1 encephalitis, which has been lacking until now.^26,49 This finding is especially noteworthy when considering that, in principle, patients who are more difficult to treat are more likely to receive second-line therapy. Despite this presumed severity bias, which we would expect to negatively confound the efficacy of second-line IT, the positive effect seems to remain.

The prognosis of the disease is further complicated by the occurrence of relapses. We observed clinical relapses in 49 of 188 patients (26%), a rate within the mid-range of the spectrum of 14%–41% suggested by existing literature.^8,11,17,50 However, the absence of a standardized definition and reliable biomarkers poses a fundamental challenge in determining what constitutes a relapse. Ambiguity is introduced when attempting to attribute seizures to an autoimmune or other, e.g., structural causes, because fluctuations after long periods of seizure freedom are also observed in patients with chronic epilepsy. Besides, the advanced age at disease onset also enables neurodegenerative processes to contribute to cognitive decline.

The main limitation of our study is its retrospective design, which must be carefully considered when interpreting our findings. However, conducting a prospective study would be highly challenging because of the rarity of the disease and its frequently prolonged and variable course. Yet, the multicenter study design and the large cohort size strengthen the robustness of our results.

In addition, seizure occurrence and freedom largely depend on self-reporting, while cognitive dysfunction is assessed based on physician impressions. Systematic bias may also arise from differing therapeutic strategies across cohorts and centers; e.g., second-line IT was more frequently used in the French cohort. Furthermore, dosage and duration of the corticosteroid therapy applied was not assessed, thereby preventing an evaluation of the effects of different regimens on outcome parameters. Unfortunately, we were unable to determine whether the use of specific ASMs was associated with faster seizure remission because detailed information on the individual ASMs administered was not available for all patients. Our CSF and serum biomarker analysis were likely underpowered because of limited biomaterial and heterogeneous sampling time points, and further studies and standardized sample collections are needed.

In summary, our study, consisting of a large, multinational sample with extended follow-up periods, may provide an accurate long-term estimate of the proportion of patients with AEAE secondary to anti-LGI1 encephalitis at 5.9%. Thus, most (>90%) of the patients do not go on to fulfil the criteria for AEAE. The ROSR of 20% (95% CI 11%–28%) 5 years after remission in patients with anti-LGI1 encephalitis was even lower than that of patients with a single unprovoked seizure. Consequently, the diagnosis of epilepsy is inappropriate in most cases upon seizure remission, and ASM withdrawal should be considered. We identified younger age at disease onset and persistent cognitive dysfunction as factors associated with AEAE. It is further important to note that our study supports the notion that the use of second-line IT, especially rituximab, could translate to faster seizure remission.

Glossary

AEAE

AIE-associated epilepsy

AIE

autoimmune encephalitis

ASM

antiseizure medication

FBDS

faciobrachial dystonic seizures

HS

hippocampal sclerosis

IQR

interquartile range

IT

immunotherapy

LGI1

leucine-rich glioma-inactivated 1

mRS

modified Rankin Scale

NfL

neurofilament light chain

ROSR

risk of seizure recurrence

Appendix. Coinvestigators

Coinvestigators are listed at Neurology.org.

Author Contributions

T. Baumgartner: 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. M. Freyberg: 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. L. Campetella: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. Y. Crijnen: major role in the acquisition of data. J. Dargvainiene: major role in the acquisition of data; analysis or interpretation of data. C. Behning: analysis or interpretation of data. C.G. Bien: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design. A. Rada: major role in the acquisition of data. H. Prüss: major role in the acquisition of data. R. Rössling: major role in the acquisition of data. S. Kovac: major role in the acquisition of data. C. Strippel: major role in the acquisition of data. F.S. Thaler: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. K. Eisenhut: major role in the acquisition of data. J. Lewerenz: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. F. Becker: major role in the acquisition of data. R. Reinecke: major role in the acquisition of data. M.P. Malter: major role in the acquisition of data. K-W. Sühs: major role in the acquisition of data. S.C. Tauber: major role in the acquisition of data. F. Von Podewils: major role in the acquisition of data. N. Melzer: drafting/revision of the manuscript for content, including medical writing for content. K-P. Wandinger: major role in the acquisition of data. R-A-M. Fernandez Ceballos: major role in the acquisition of data; analysis or interpretation of data. J. Kuhle: major role in the acquisition of data; analysis or interpretation of data. K. Berger: major role in the acquisition of data; analysis or interpretation of data. T. Bauer: major role in the acquisition of data. T. Rüber: major role in the acquisition of data. A. Racz: major role in the acquisition of data. A.J. Becker: drafting/revision of the manuscript for content, including medical writing for content. J. Pitsch: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data. G. Kuhlenbäumer: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. S. Muñiz-Castrillo: major role in the acquisition of data. J. Honnorat: major role in the acquisition of data. M.J. Titulaer: 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. F. Leypoldt: 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. R. Surges: 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 preparation of this article and the research presented herein were in part supported by grants from the German Federal Ministry of Education and Research (CONNECT-GENERATE grant nos. 01 GM1908A and 01 GM2208A). The French data have been collected within the BETPSY (Biomarkers in autoimmune encephalitis and paraneoplastic neurological syndromes) project, which is supported by a public grant overseen by the French national research agency (Agence nationale de la recherche (ANR)), as part of the second “Investissements d'Avenir” program (reference ANR-18-RHUS-0012). The Dutch research (Y.S.C., M.J.T.) was supported by a grant from EpilepsieNL, project number NEF 19-08. Moritz Freyberg's work was supported by a BONFOR SciMed grant (2021-4-26).

Disclosure

T. Baumgartner, M. Freyberg, L. Campetella, Y.S. Crijnen, J. Dargvainiene, C. Behning, C.G. Bien, A. Rada, H. Prüss, R. Rößling, S. Kovac, C. Strippel, F.S. Thaler, K. Eisenhut, J. Lewerenz, F. Becker, R. Reinecke, and M. Malter report no disclosures. K-W Sühs received honoraria for lectures or travel reimbursements for attending meetings from Biogen, Merck, Mylan, Roche, Bavarian Nordic, Viatris, and Bristol-Myers Squibb, as well as research support from Bristol-Myers Squibb. S.C. Tauber reports no disclosures. FV Podewils has received personal fees as a speaker or for serving on advisory boards from Angelini, Arvelle, Bial, Desitin Arzneimittel, Eisai, Jazz Pharmaceuticals, UCB Pharma, and Zogenix. N. Melzer, K-P. Wandinger, R.A. M. Fernandez Ceballos, J. Kuhle, K. Berger, T. Bauer, T. Rüber, A. Racz, A.J. Becker, J. Pitsch, G. Kuhlenbäumer, and S. Muñiz-Castrillo report no disclosures. F. Leypoldt discloses speaker honoraria from Grifols, Argenx, and Roche; travel funding from Grifols; and service on advisory boards for Roche and Argenx. MJ Titulaer has filed a patent, on behalf of the Erasmus MC, for methods for typing neurologic disorders and cancer, and devices for use therein, and has received research funds for serving on a scientific advisory board of AmGen and for consultation at Guidepoint Global LLC and UCB, royalties from UpToDate Inc., and an unrestricted research grant from Euroimmun AG and from CSL Behring. R. Surges reports no disclosures. Go to Neurology.org/NN for full disclosures.