Trajectories of long‐term survival after status epilepticus

Abstract

Objectives

To investigate the trajectories of clinical characteristics and prognostic factors among long‐term survivors of status epilepticus (SE), given the high mortality during acute hospitalization and in subsequent years.

Methods

Adult patients (≥18 years of age) with first‐time, non‐anoxic SE were identified and included from University Hospitals in Oslo (Norway), Odense (Denmark), Frankfurt, and Marburg (Germany). Demographics, etiology, comorbidities, and seizure characteristics were assessed. Poisson regression was used to model mortality rates over the follow‐up period.

Results

Between 2001 and 2017, we included 1306 patients (median follow‐up: 17.4 months). The estimated cumulative proportion surviving at 3, 12, 24, and 36 months were 94.0%, 73.0%, 51.1%, and 39.5%, respectively, with a similar increase in mortality after discharge across all cohorts. Daily mortality peaked during the first 150 days; mortality trajectories differed depending on etiology, SE duration, and age. The clinical characteristics of survivors changed during long‐term follow‐up; long‐term survivors (>36 months) were younger, had shorter SE durations, and had different underlying etiologies. The relative impact of different prognostic factors on the daily mortality shifted during long‐term follow‐up. Although most established prognostic factors strongly influenced in‐hospital mortality, the relative impact of SE duration, comorbidities, and remote symptomatic etiologies first peaked after 6 months.

Significance

The optimal time point to assess survival in the acute phase is at 6 months, whereas evaluating survival after 2.5 years provides reliable estimates of long‐term mortality. Assessing SE survival at discharge underestimates the impact of remote symptomatic etiologies and duration of SE on long‐term survival.

Key points.

• Long‐term mortality exceeds in‐hospital mortality by 300%–400%.

• Mortality after status epilepticus stabilizes after ~6 months.

• The optimal time point to assess survival in the acute phase is at 6 months.

• The dynamics of mortality and the trajectories of long‐term survival differ substantially between etiologies.

• Duration of status, remote symptomatic etiologies, and comorbidities impact long‐term but not short‐term survival.

1. INTRODUCTION

Status epilepticus (SE) is a common neurological emergency associated with high mortality and morbidity in adults. ¹ , ² , ³ , ⁴ Treatment may comprise intensive care, sometimes including general anesthesia associated with potential complications. Around 15% of patients die during the acute phase, often due to withdrawal from therapy due to continuous epileptic activity or due to the underlying disease. ⁵ , ⁶ , ⁷ Accurate outcome prediction remains difficult despite substantive efforts in recent years. ⁸ , ⁹ Although several key prognostic factors for in‐hospital mortality are established and incorporated into prognostic scores, ¹⁰ , ¹¹ , ¹² their application in clinical practice remains challenging. ¹³ The etiological heterogeneity of SE complicates exact prognostication because an absence SE after, for example, a pause of medication will hardly affect survival, whereas SE associated with anoxia almost always leads to a dismal outcome. ⁸ , ¹⁴ High long‐term mortality presents a second major challenge. Several cohorts with available long‐term follow‐up reported unequivocally an increased long‐term mortality as compared to in‐hospital mortality. ⁴ , ¹³ , ¹⁵ , ¹⁶ Recent epidemiological evidence from a retrospective hospital cohort suggests that the causes of death and associated risk factors may differ depending on the time point assessed. ¹⁷

Most retrospective and large prospective cohorts studied in‐hospital mortality, which is easily amenable without the risk of follow‐up losses. ⁵ , ¹⁴ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ In these studies, etiology, electroencephalography (EEG) changes, history of epilepsy, worst seizure type, comorbidity, age, and so on, have been identified as major prognostic factors. In contrast, the duration of seizures or time in SE, as well as treatment in the intensive care unit (ICU) have not been reported consistently as risk factors for in‐hospital mortality, unless seizures were very brief or patients had super‐refractory SE, defined as the failure of at least one cycle of general anesthesia. ²²

The reported risk factors vary depending on the chosen observation period and endpoint. These differences were most pronounced for SE duration that hardly affects in‐hospital mortality but predicts 3‐month survival and functional outcomes. ²³ In line, we recently identified the duration of SE as a major risk factor for long‐term survival and functional outcomes at discharge, allowing for predicting long‐term survival in different cohorts of patients with SE due to non–brain‐damaging etiologies. ¹² These findings corroborate recent magnetic resonance imaging (MRI) studies indicating that peri‐ictal MRI changes were not associated with in‐hospital mortality but are linked to duration, functional outcomes, and long‐term survival. ²⁴ , ²⁵ Etiologies may also influence outcomes differently at different time points. This effect is particularly evident in patients with glioblastoma, who rarely survive more than 2 years, with SE being a negative prognostic factor. ²⁶ , ²⁷

In summary, the available evidence suggests that different risk factors influence survival differently depending on the time point and endpoint of assessment. Understanding the temporal dynamics and trajectories of prognostic factors is crucial for both treatment decisions and the reconciliation of conflicting data in the literature. Therefore, we used a large, multinational cohort to understand the trajectories of long‐term survivors after SE.

2. MATERIALS AND METHODS

2.1. Approvals

The Danish cohort received approval from the Danish Health Authorities, who also granted a waiver for the informed consent requirement; under Danish legislation, approval from an ethics committee was not required. The evaluation of German patients was part of a study on SE outcomes, registered with the German Clinical Trials Register and approved by local ethics committees in Frankfurt and Marburg, Germany. The Norwegian data were collected as part of a quality assessment study, approved by the Data Protection Official, with permission to publish the results in the public interest.

2.2. Description of the cohorts

All cohorts have been described previously, and the information specified in the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines can be found in the source publications. ¹² , ²⁸ , ²⁹ , ³⁰ , ³¹ Inclusion criteria required patients to be 18 years or older and to have a diagnosis of first‐time SE, defined as convulsive seizures lasting longer than 5 min, non‐convulsive seizures lasting longer than 10 min, or non‐convulsive or possible non‐convulsive SE demonstrated on EEG according to the latest 2015 International League Against Epilepsy (ILAE) Classification of SE and the Salzburg criteria, respectively. ³² , ³³ All kinds of SE were included irrespective of treatment response; patients with later SE episodes in addition to the first SE evaluated here were not excluded from the study. The diagnosis of SE was ascertained by analysis of the patients’ records of all patients and by reevaluation of EEG. ³⁰ , ³⁴ Patients with anoxic brain damage were excluded. Identical inclusion and exclusion were applied to all cohorts to secure homogeneity. Completeness of follow‐up was ascertained via electronic or postal contact with the civil registration offices/systems as described in the cited references.

The Danish cohort comprised patients diagnosed and treated at Odense University Hospital between January 1, 2008 and December 31, 2017. These patients were identified retrospectively from referrals for acute EEG studies and/or based on International Classification of Diseases, Tenth Revision (ICD‐10) codes at discharge as described in. ⁴

The German cohort included patients diagnosed at Frankfurt University Hospital and the University Hospital Marburg between January 1, 2011 and December 31, 2017. All patients with SE were included, and details of this subset have been described previously. ²⁸ , ²⁹

The Norwegian cohort consisted of patients diagnosed at Oslo University Hospital between January 1, 2001 and December 31, 2017. Patients with SE were identified retrospectively using ICD‐10 codes at discharge, with data covering the period from 2001 to 2017. Information on these patients has been reported elsewhere. ³⁰ , ³⁵

2.3. Collection of outcome variables and covariates

The electronic medical records were analyzed retrospectively by individual raters, with the local principal investigator making the final assessment in cases of uncertainty (K.H., A.S., S.K., and C.P.B.). If uncertainty remained, the data point remained missing. We assessed demographic characteristics, the duration of SE, semiological traits, all components (level of consciousness, history of seizures, worst seizure type) of the Status Epilepticus Severity Score (STESS), ¹⁰ and components of the Epidemiology‐Based Mortality Score in Status Epilepticus (EMSE), ³⁶ including the Charlson Comorbidity Index (CCI). For all patients, etiology was assessed. Inspired by the ILAE's recommendations, ³⁷ we further subcategorized aetiologies as follows:

• “Other causes”: toxic causes (such as alcohol, recreational drugs, and prescription medications) and metabolic disturbances.

• “Epilepsy”: withdrawal or low levels of antiseizure medications (ASMs), known epilepsy with unprovoked seizures.

• “Acute symptomatic”: SE due to acute ischemic or hemorrhagic stroke, infectious, or autoimmune meningitis/encephalitis.

• “Remote symptomatic”: SE as first manifestation of a symptomatic epilepsy due to pre‐existing structural damage (e.g., post‐stroke).

• “CNS tumors”: all kinds of SE episodes that could be unequivocally assigned to a CNS tumor.

Scores were approximated retrospectively based on the available information in the electronic medical records; STESS and EMSE were assessed at admission; the Age‐conciousness‐duration (ACD) score at the time point of seizure cessation. Semiological traits were classified according to the recommendations of the ILAE. ³⁸

The study's primary endpoint was all‐cause mortality/survival. For the German cohort, survival status was determined through regular follow‐ups with local registration offices in the patients' areas of residence. The Danish cohort's follow‐up was complete due to the linkage of the electronic patient files with the Danish Civil Registry. In Norway, survival data for all patients were available due to the linkage between the hospital's electronic medical records and the Norwegian National Population Register. Expected mortality rates of the age groups covered by our cohort were estimated using data from Danish normal population controls that was retrieved from Statistics Danmark on Feb 18, 2025. The causes of death were determined retrospectively in the Danish cohort based on available hospital records and the treating physicians’ evaluations. If data were inconclusive or lacking, the data point was left blank. Epilepsy‐releated death was defined as death due to new SE or death in association with an observed epileptic seizure.

2.4. Statistical analyses

Statistical analyses were performed using R version 4.4.1, relying on packages popEpi (version 0.4.12), Epi (2.55), survival (3.5–7), and SPSS V21. Baseline data are summarized as frequencies and percentages for discrete data and as median and interquartile range (IQR) for continuous data. We used chi‐square test/Fisher exact test for categorical data and Kruskal–Wallis test for numeric data as appropriate and without correct for multiple testing. For survival analysis, the log‐rank test was used. Time‐to‐event data consists of time since SE diagnosis until death or end of follow‐up. Poisson regression including natural cubic splines was used to model mortality incidence rates over follow‐up time, which was split into periods of 10 days. The explanatory variables were age, CCI, duration of SE (log transformed), level of consciousness, history of seizures, worst seizure type, and etiology. Analyses were inspired by Carstensen (2022). ³⁹

The analyses are summarized by plotting the estimated mortality incidence rate (IR) (simple analysis) or the incidence rate ratios (IRRs) as functions of time with 95% confidence intervals (CI). If the lower bound of the CI was above the reference, statistical significance was assumed. Details of the statistical analyses are given in the supplementary methods.

3. RESULTS

3.1. Patient demographics and long‐term survival

The demographics of patients included in the different cohorts are given in Table 1, whereas Figure 1A presents a Kaplan–Meier survival curve for the three cohorts. After accounting for censored patients, the estimated cumulative survival proportions were 94.0%, 73.0%, 51.1%, and 39.5% at 3, 12, 24, and 36 months, respectively. Loss to follow‐up was infrequent (n = 10, .8%). Figure 1B illustrates that the daily mortality rate declined rapidly during the first 100 days following SE and then decreased steadily throughout the observation period, although it remained at a high level. Demographic details, including median age and survival rates, differed between the cohorts (Table 1), with clinically relevant differences in the rate of older patients and of acute‐symptomatic etiologies that were lower in the Norwegian cohort. The relative mortality incidence after discharge (proportion of patients who died in hospital compared to patients dying after discharge) increased; however, at similar rates in three cohorts, as shown in Figure 1C.

TABLE 1.

Clinical characteristics of the cohort.

	Denmark (n = 271)	Germany (n = 906)	Norway (n = 139)
	N (%)	N (%)	N (%)
Median age at onset, years (IQR)	68 (58–78)	70 (55–79)	54 (38–71)
Median follow‐up, days (IQR)	534 (66–1221)	421.5 (41–1098)	2040 (540–4140)
Median duration of SE, h (IQR)	71 (23–157)	47 (9–189)	24 (3–120)
Median Charlson Comorbidity Index (IQR)	4 (2–6)	3 (1–4)	1 (0–2)
Median ACD score (IQR)	9 (8–11)	9 (6–11)	7 (5–9)
Median STESS (IQR)	3 (2–5)	3 (2–4)	2 (1–3)
Etiology
Epilepsy/poor adherence	6 (2.3)	16 (1.8)	14 (10.1)
Acute symptomatic	76 (29.1)	248 (27.4)	10 (7.2)
Remote symptomatic	81 (31.0)	233 (25.7)	32 (23.0)
Toxic causes	14 (5.4)	30 (3.3)	5 (3.6)
Metabolic disturbances	15 (5.7)	17 (1.9)	1 (.7)
CNS tumors	40 (15.3)	177 (19.5)	16 (11.5)
Epilepsy/unprovoked SE	9 (3.4)	132 (14.6)	48 (34.5)
Unknown	20 (7.7)	53 (5.8)	13 (9.4)
Age above 65 years and above
No	99 (37.9)	362 (40.0)	93 (66.9)
Yes	162 (62.1)	544 (60.0)	46 (33.1)
Level of consciousness at admission
Awake	120 (46.0)	494 (54.5)	59 (42.4)
Comatose	141 (54.0)	412 (45.5)	80 (57.6)
History of seizures
No	146 (55.9)	368 (40.6)	100 (71.9)
Yes	115 (44.1)	538 (59.4)	39 (28.1)
Worst seizure type
Generalized tonic–clonic	84 (32.2)	476 (52.8)	33 (23.7)
Focal	95 (36.4)	326 (36.1)	87 (62.6)
NCSE in coma	82 (31.4)	100 (11.1)	19 (13.7)
Patient having survived 3 months	185 (68.2)	598 (66.0)	118 (84.4)
Patients having survived 12 months	142 (52.4)	471 (52.0)	107 (77.0)
Patients having survived 24 months	113 (41.7)	339 (52.0)	96 (69.1)
Patients having survived 36 months	61 (22.5)	228 (25.2)	91 (65.5)

FIGURE 1.

Long‐term survival following status epilepticus. (A) Kaplan–Meier curves illustrating long‐term survival following status epilepticus (SE) in Germany, Denmark, and Norway. Age‐matched Danish normal population controls were used to illustrate the expected mortality of the cohort (dotted line). (B) The daily mortality rates per 100 person days following admission due to SE is presented. (C) The relative increase in mortality during follow‐up as compared to in‐hospital mortality (= 100%) is given. (D–F) The proportion of total deaths that have occurred by a given time point is shown for patients with (D) different etiologies, (E) age categories, and (F) duration of seizures. The small inslet shows the cumlative incidence of mortality during the first 2 months.

3.2. Dynamics of mortality

We examined three main determinants of long‐term survival to assess long‐term mortality dynamics. The evolution of the cumulative incidence of mortality differed by etiology during long‐term follow‐up (Figure 1D); 80% of patients who died from CNS tumors or acute causes died within 313–335 days. Of all patients who died during follow‐up and had SE due to known epilepsy, remote symptomatic causes, or other causes, 80% died within 1380, 837, and 884 days, respectively. In contrast, the effect of age was stable from the beginning and changed only slightly (Figure 1E). The duration of SE also influenced the dynamics of mortality. Although essentially all patients without successful cessation of seizures at discharge died within 2 months, the cumulative mortality rate of patients with very long SE episodes (>149 h) gradually increased, with differences becoming apparent only during long‐term follow‐up (Figure 1F).

3.3. Clinical characteristics of survivors

Table 2 shows the characteristics of all patients at seizure onset and of survivors at different time points (onset, discharge, and after 3, 12, 24, and 36 months). The pattern of surviving patients after SE changed substantially. Notably, the median duration of SE fell from 48 h (all patients, 95% CI for median: 44–62 h) to 32 h for those surviving at least 36 months (95% CI for median: 24–41 h, p < .001 Kruskal–Wallis test). The main driver was the high mortality of patients with very long SE (Figure 2A). Surviving patients were younger, with a median age of 59 years at onset (95% CI for median: 57–63 years) compared to 69 years in the initial cohort (95% CI for median: 68–71 years, p < .001 Kruskal–Wallis test), mainly due to a relatively increased mortality of patients of 65 years and older (Figure 2B). A similar trend was seen for comorbidity, with the median CCI decreasing from 3 to 2 (Table 2). In contrast, differences in established seizure‐related prognostic factors such as history of epilepsy, seizure semiology, or consciousness at onset, were less pronounced (Table 2). The distribution of underlying etiologies among survivors remained largely unchanged in the long‐term compared to the original cohort, except for a decrease in patients with CNS tumors and a relative increase in patients with SE due known epilepsy (Figure 2C). Data on the causes of death were available for subgroups from the Danish cohort. Death due to new SE occurred in the first months after discharge (Figure 2D), high modified Rankin scale at discharge was associated with death not due to the underlying etiology (Figure S1). Long‐term survival did not improve between 2011 and 2017 (Figure 2E).

TABLE 2.

Change of clinical characteristics in survivors after status epilepticus.

	Alive at
	Onset N =1306	At discharge n = 1088	3 months N = 901	12 months N = 720	24 months N = 548	36 months N = 380	p‐value
	Median (IQR)/N (%)	Median (IQR)/N (%)	Median (IQR)/N (%)	Median (IQR)/N (%)	Median (IQR)/N (%)	Median (IQR)/N (%)
Median age at onset, years (IQR)	69 (54–78)	66 (52–77)	64 (51–75)	61 (49–74)	60 (47–73)	59 (45–72)	<.001 b
Median follow‐up, days (IQR)	523 (49–1327)	734 (180–1571)	980 (480–1810)	1206.5 (740.5–1984.5)	1553 (1053–2215.5)	1920 (1478–2483.5)	<.001 b
Median duration of SE, h (IQR)	48 (11–179)	45 (7–164)	41 (7–161)	36 (6–144)	33 (5–144)	32 (5–143)	.01 b
Median Charlson Comorbidity Index (IQR)	3 (1–5)	2 (1–4)	2 (1–4)	2 (1–4)	2 (1–4)	2 (0–3)	<.001 b
Median ACD score (IQR)	9 (6–11)	8 (6–10)	8 (6–10)	8 (6–10)	8 (6–10)	7 (5–10)	<.001 b
Median STESS (IQR)	3 (2–4)	3 (2–4)	3 (1–4)	2 (1–4)	2 (1–3)	2 (1–3)	<.001 b
Etiology
Epilepsy/poor adherence	36 (2.8)	36 (3.3)	35 (3.9)	31 (4.3)	29 (5.3)	25 (6.6)	<.001 a
Acute symptomatic	334 (25.6)	244 (22.4)	196 (21.8)	164 (22.8)	130 (23.7)	82 (21.6)
Remote symptomatic	346 (26.5)	294 (27)	247 (27.4)	203 (28.2)	155 (28.3)	110 (28.9)
Toxic causes	49 (3.8)	48 (4.4)	42 (4.7)	36 (5)	30 (5.5)	24 (6.3)
Metabolic disturbances	33 (2.5)	28 (2.6)	27 (3)	21 (2.9)	18 (3.3)	10 (2.6)
CNS tumors	233 (17.8)	194 (17.8)	134 (14.9)	74 (10.3)	46 (8.4)	32 (8.4)
Epilepsy/unprovoked SE	189 (14.5)	175 (16.1)	160 (17.8)	139 (19.3)	105 (19.2)	74 (19.5)
Unknown	86 (6.6)	68 (6.3)	60 (6.7)	52 (7.2)	35 (6.4)	23 (6.1)
Level of consciousness at admission
Awake	673 (51.5)	589 (54.2)	501 (55.6)	398 (55.3)	292 (53.3)	199 (52.4)	.313 a
Comatose	633 (48.5)	498 (45.8)	400 (44.4)	322 (44.7)	256 (46.7)	181 (47.6)
History of seizures
No	614 (47)	533 (49)	451 (50.1)	359 (49.9)	285 (52)	205 (53.9)	.102 a
Yes	692 (53)	554 (51)	450 (49.9)	361 (50.1)	263 (48)	175 (46.1)
Worst seizure type
Generalized tonic–clonic	593 (45.5)	500 (46.1)	404 (44.9)	311 (43.3)	227 (41.6)	142 (37.5)	<.001 a
Focal	508 (39)	446 (41.1)	384 (42.7)	323 (45)	255 (46.7)	206 (54.4)
NCSE in coma	201 (15.4)	139 (12.8)	111 (12.3)	84 (11.7)	64 (11.7)	31 (8.2)

^a Chi‐square test.

^b Kruskal–Wallis test.

FIGURE 2.

The clinical characteristics of the initial status epilepticus (SE) population and long‐term survivors change over time. (A) The changes in the relative proportions of patients surviving after having SE with a duration of 1–6 h, 7–149 h, and >149 h. Patients with brief SE <1 h are not included in the analysis (censored patients with insufficient follow‐up were excluded, *p < .001, chi‐square test comparing non‐survivors with survivors). (B) The proportion of patients 65 years of age or older at onset within the cohorts of surviving patients is given (censored patients with insufficient follow‐up were excluded, p < .001, chi‐square test comparing non‐survivors with survivors). (C) The changes in the relative proportions of different etiologies in patients alive at onset, discharge, and after 3 to 36 months are illustrated (censored patients with insufficient follow‐up were excluded, p < .001, chi‐square test comparing non‐survivors with survivors). (D) Temporal distribution of epilepsy‐ and non–epilepsy‐related deaths during follow‐up depending on the Epidemiology‐Based Mortality Score in Status Epilepticus (EMSE) score at onset. (E) Long‐term survival of patients treated for first time SE in the years 2011 to 2017 is illustrated using Kaplan–Meier curves without statistically significant differences (log‐rank test).

3.4. Temporal change of the prognostic impact of established prognostic factors

The changing clinical characteristics of survivors compared to the clinical characteristics of the cohorts at diagnosis suggest that different risk factors affect survival and mortality at different time points. We therefore calculated the IRRs as a function of time for a panel of prognostic factors available in all three cohorts. Age at onset had a consistent impact throughout the follow‐up period, with a slight peak during the acute phase and again at the end of the observation period (Figure 3A). It is important to note that the prognostic impact of the CCI and the duration of SE peaked at around 400 and 150 days, respectively (Figure 3B,C). Seizure characteristics, level of consciousness, and history of epilepsy were of prognostic importance during the acute phase (Figure 3D–F). The relative impact of different etiologies compared to the entire cohort unveiled a substantial effect of acute symptomatic etiologies in the acute phase, a very strong effect of CNS tumors on survival in the subacute phase, and a consistent and stable negative effect of remote symptomatic seizures during long‐term follow‐up, particularly after ~1 year (Figure 4).

FIGURE 3.

Changes in the impact of established prognostic factors on daily mortality. (A–F) The gray horizontal dotted line signifies a mortality incidence rate ratio of 1. The vertical axis is logarithmic. The graphs display the incidence rate ratio and 95% confidence intervals (dashed lines) following admission due to SE for (A) age, (B) Charlson Comorbidity Index, (C) duration of status epilepticus, (D) level of consciousness at admission (as defined by the Status Epilepticus Severity Score [STESS]), (E) history of seizures, and (F) worst seizure type at admission (STESS categories).

FIGURE 4.

Changes in the impact of the triggering etiology on daily mortality. The gray horizontal dotted line signifies mortality incidence rate ratio of 1. The vertical axis is logarithmic. The graphs display the incidence rate ratio and 95% confidence intervals (dashed lines) for the eight etiological categories compared to the marginal mortality incidence rate.

4. DISCUSSION

The most important finding of our study is that the presence of various clinical factors is associated with increased mortality at different time points to different degrees, leading to significant changes in the composition of the group of survivors after SE. This is particularly important for three factors that strongly affect survival in the first 1.5 years after the SE episode: duration of the SE, comorbidity, and remote symptomatic causes of SE that hardly affect survival in the acute phase but are associated with higher long‐term mortality. Thus, the impact of these factors is underestimated when looking at only in‐hospital mortality because patients with very long SE, many comorbidities, or patients with epilepsy due to previous brain damage will have a higher risk of dying after discharge. Although the link between comorbidity, previous brain damage, and survival is obvious, the association between the duration of seizures and the long‐term mortality is far less clear. Given the strong associations of long‐term survival with seizure duration, new neurological deficits at discharge, ¹² and functional outcomes, ²³ we interpret the delayed effect of seizure duration on survival as likely reflecting SE‐induced neurological damage. Preliminary data suggesting that patients discharged in poor condition will likely die for reasons other than the underlying conditions appear to support this conclusion. However, these retrospective data are substantially biased toward patients with a need for follow‐up in a university hospital (mainly cancer patients) and prospective studies are need to confirm our data. Death due to epilepsy‐related causes, especially new SE, appears to be rare and happened during the first 6 months.

Our study allows for reconciling the conflicting observations in the literature regarding the prognostic impact of SE duration on survival. Consistent with previous reports, the impact of seizure duration on short‐term survival is low ²² , ⁴⁰ as its impact peaks around 6 months. Short‐term analyses could be influenced by very mild episodes of SE and by patients dying of SE, both of which could weaken correlations. In contrast and in line with our data, publications reporting on later outcomes ¹⁶ or functional outcomes ²³ have found significant associations.

The most important implication may be for clinical practice. Although new tools like the ACD score aid in prognosticating patients with SE due to non–brain‐damaging etiologies, ¹² the combination of acute cerebral pathology with SE complicates individual prognostication. Our interpretation of our results is that a two‐step approach to prognostication might be useful. If a patient with acute brain disease has a reasonable chance of survival, the duration of SE (estimated using the ACD score) and comorbidities will determine intermediate‐term survival later. Although our data do not allow individualized prognostication, understanding the temporal course of the different prognostic factors will assist neurologists in making these difficult treatment decisions at the bedside. In this context, it is important to acknowledge the CIs that were narrow for common conditions and large for more rare diseases with variable outcome (e.g., metabolic disease).

Cohorts reporting long‐term survival are rare, and outcomes depend largely on the patients included. It is important to note that children generally have substantially better survival rates compared to adults and the elderly. ¹⁶ , ⁴¹ In this multicenter European study, we observed demographic differences between the cohorts. Patients from the Oslo University Hospital had fewer acute symptomatic etiologies and more younger patients, which is not surprising given its role as highly specialized referral center for Norway in addition to acute care. In contrast, the Danish and German hospitals primarily provide acute care in their respective catchment area with almost identical survival curves. Nevertheless, the ratio of mortality after discharge and in‐hospital mortality was similar among the different centers and increased to 3–4 after 1000 days of follow‐up (Figure 1C). This finding supports our own long‐term data from a population‐based study, ⁴ data from a Finish study, ¹⁵ and high‐quality epidemiological data from New Zealand. In Auckland, the authors reported a 30‐day mortality of 4.6% ⁴⁰ that increased to 14.9% after 2 years in a large mixed cohort of pediatric and adult patients. ¹⁶ For patients 25–100 years of age, the respective numbers from the same study were 8.1% and 30.7%, which is similar to our results. ¹⁶ , ⁴⁰

Important aspects not addressed extensively in our study are the causes of death that comprise epilepsy‐related causes (mainly recurrent SE), death facilitated by the neurological damage induced by SE, complications of treatment, and the underlying disease. ⁴² , ⁴³ Thus, SE does not have a unidirectional effect on death, for example the underlying etiology leading to death may also predispose patients to SE, and although the causes of death vary between patients, the major causes of death are usually interdependent. Our data complement this knowledge by showing that epilepsy‐related death requiring hospitalizations are rare and that only age, CCI, and having a malignant brain tumor were prognostic factors during the entire follow‐up period. All other factors had variable impact on incidence mortality rates during follow‐up. For instance, the seizure‐related components of the STESS rapidly lost their prognostic yield after the acute phase, in line with STESS being designed to orient early treatment strategies. ¹⁰ The reasons for the increasing IRRs of impaired concioussness at admission and history of epilepsy after ~2 years of follow‐up are unknown. A possible explanation might be the very favorable outcome of patients with idiopathic epilepsies and essentially no excess mortality ¹² that influences the results due to a decreasing fatality rate.

In‐hospital mortality is strongly associated with withdrawal of life‐sustaining treatment, often due to the severity of the underlying neurologic disease or the inability to liberate patients from anesthetic infusion without recurrence of seizures, ⁶ corroborating the strong prognostic impact of acute symptomatic causes of SE directly after diagnosis. After discharge, comorbidities, age, recurrent status, and previous and newly acquired neurological deficits after SE are likely contributors to the high all‐cause mortality in the long term. However, more detailed studies are needed to further understand the causes of death after discharge.

Cumulative evidence clearly indicates that the majority of patients die after discharge from hospital, making in‐hospital mortality an important but insufficient endpoint for clinical studies. Given that the incidence of mortality stabilizes ~6 months after discharge, when looking at the estimated survival for all patients, we propose using a 6‐month survival as an evidence‐based primary outcome parameter for clinical trials and epidemiological studies assessing SE‐associated mortality. For long‐term mortality, a 30‐month follow‐up appears optimal, although final assessments are difficult because the survival curves do not plateau and excess mortality persists throughout the follow‐up period.

An undisputable strength of our study is the large size of the cohort exceeding the largest national registries to date. ⁴⁴ Combined with the complete follow‐up of the Scandinavian cohorts and almost complete long‐term follow‐up of the German cohorts, selection bias from unknown external factors appears very unlikely. On the other hand, the retrospective design and the merging of the cohorts did reduce the granularity of the available data, and we had to restrict our analyses to the prognostic factors assessed in the EMSE, ³⁶ STESS, ¹⁰ and ACD ¹² scores that were available across all cohorts, which is a limitation of our study. For instance, data on the promptness of therapy initiation, new neurological deficits, timely administration of etiological treatment, and so on, are lacking despite their obvious importance for short‐ and long‐term outcomes. In addition, the causes of death shown in Figure S1 are likely biased due to overrepresentation of patients with brain cancer and available follow‐up at the hospital and underpresentation of elderly patients dying in a nursing home. Consequently, we cannot draw conclusions about factors other than those included in our study, and the dynamics of the prognostic impact of EEG or MRI changes on long‐term outcomes remains to be determined.

In conclusion, our study highlights the variability of the influence of different risk factors over time, which may explain contradictory results of epidemiological studies. Age, acute symptomatic etiologies, and other established risk factors for in‐hospital mortality mainly predict short‐term survival, whereas remote symptomatic etiologies, comorbidities, and the duration of SE become increasingly important in the follow‐up period. Assessing SE survival at discharge underestimates the impact of remote symptomatic etiologies, comorbidities, and duration of SE on long‐term survival. The optimal time point to assess survival in the acute phase is at 6 months, whereas evaluating survival after 2.5 years provides reliable estimates of long‐term mortality.

AUTHOR CONTRIBUTIONS

C.D.C., S.B.K., A.S., and C.P.B. contributed to the conception and design of the study. L.B.U., K.H., E.T., F.R., A.S., S.K., and C.P.B. contributed to acquisition and analysis of data. C.D.C., S.B.K., and C.P.B. contributed to drafting the text and preparing the figures.

CONFLICT OF INTEREST STATEMENT

Nothing to report. Conflicts of interest that are not relevant to this research activity are reported on the conflict of interest forms.

ETHICS STATEMENT

We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Supporting information

Figure S1.

Data S1.

Cornwall CD, Kristensen SB, Ulvin LB, Heuser K, Taubøll E, Strzelczyk A, et al. Trajectories of long‐term survival after status epilepticus. Epilepsia. 2025;66:2790–2802. 10.1111/epi.18428

DATA AVAILABILITY STATEMENT

The anonymized dataset is made available to qualified researchers by request.