Address correspondence and reprint requests to Dr. F. Rosenow at Department of Neurology, Philipps-University Marburg, Rudolf-Bultmann-Str. 8, 35033 Marburg, Germany. E-mail: firstname.lastname@example.org
Summary: Purpose: The aim of this study was to determine the long-term case fatality of patients with a first episode of status epilepticus (SE group) of cerebrovascular etiology, as compared with that in acute stroke patients without SE (AS group).
Methods: Patients with SE who had been prospectively admitted to an epidemiologic study were retrospectively compared with a cohort of patients from the local stroke registry. The main outcome end point was overall survival. Survival curves were generated according to the Kaplan–Meier method and compared by using the log-rank test. An extended Cox model was used to examine the impact of patient group on the risk of death. Covariates considered potential confounders included age at diagnosis, sex, type of stroke, affected hemisphere, and localization of lesions.
Results: Of 166 patients who entered the study, 93 patients were in the SE group, and 73 patients were in the AS group; 53 SE patients and 35 AS patients died during the study. Patient group (SE vs. AS) showed no significant impact on survival (p = 0.0832) in univariate analysis. In contrast, the results from a multivariable analysis suggest that after 6 months, patients with SE were at about twice the risk of death as were patients with AS [hazard ratio of 2.12 with 95% confidence interval, 1.04–4.32, p = 0.0392].
Conclusions: The occurrence of SE in patients with cerebrovascular disease indicates a high risk of death within 3 years. In contrast, the case fatality risk attributable to recurrent status or seizures is lower.
However, other studies suggest that SE might contribute to mortality. Adult patients with chronic epilepsy developing SE because of low antiepileptic drug (AED) levels were reported to have a short-term mortality of 4% (DeLorenzo et al., 1995). In a large population of patients with mild developmental disabilities, subjects with epilepsy and recent SE had the highest excess mortality (crude death rate, 8.8/1,000 person-years) as compared with controls without epilepsy (2.7/1,000) (Strauss et al., 2003). Furthermore, Waterhouse et al. demonstrated a statistically significant synergistic effect of combined injuries from SE and cerebrovascular ischemia on mortality (Waterhouse et al., 1998). Patients with acute ischemic stroke (IS) have a short-term case fatality of ∼10% at 30 days and 20–25% at 1 year (Anderson et al., 1994; Sacco et al., 1994). However, long-term case fatality is ∼50% and thus not significantly different from that of patients with intracerebral hemorrhage (ICH) (Dennis et al., 1993).
The present study compared survival in patients with the first episode of SE of vascular origin (SE group) with a series with AS, but no SE or seizures (AS group). The primary hypothesis was that SE itself is associated with increased long-term case fatality.
PATIENTS AND METHODS
Status epilepticus group
The SE group consisted of patients from the zip-code area 35—, Hessen, Germany, with 743,258 adult inhabitants (July 1, 1998), who were prospectively screened for an epidemiologic study performed by the Department of Neurology at the University Hospitals in Marburg, Germany (Knake et al., 2001). Included in this study were all patients aged 18 years or older and diagnosed with SE between July 1997 and March 2000.
SE was defined according to International League against Epilepsy (ILAE) criteria either as a single seizure lasting ≥30 min or as repeated seizures without recovery of consciousness for 30 min.
Only patients with SE caused by either remote or acute cerebrovascular events—IS and ICH—as determined by CT or MRI were included. As the SE group was recruited during a multicenter study, cranial CT scans and reports were not always available for review. As strokes can remain clinically silent and because it is frequently impossible to discern stroke-related symptoms from SE-related symptoms in patients with SE of cerebrovascular etiology, we did not attempt to differentiate between acute and remote stroke in this group. A causal relation between the cerebrovascular event and the SE was assumed if evidence was found of an acute or remote IS or ICH. Patients with subcortical arteriosclerotic encephalopathy (SAE) and patients with SE and ICH caused by arteriovenous malformations (AVMs) were excluded.
Acute stroke group
The SE group was retrospectively compared with a cohort of consecutive patients identified from the stroke unit registry of the same Department of Neurology that conducted the status study (Knake et al., 2001). Only patients diagnosed between January 1999 and March 1999 with acute IS and ICH, free of any seizures, were included. A stroke was classified as acute, if it occurred up to 7 days before admission to the hospital. Patients with SAE, transitory ischemic attacks, ICH caused by an AVM, and diagnoses other than a stroke were excluded.
The following data were collected for both groups: patient age at diagnosis, sex, type of stroke, affected hemisphere, and localization of lesions. Missing or insufficiently detailed information was considered unknown. Follow-up data, such as the date and cause of death, were obtained from medical records and through a telephone interview. If contact with the patient or a close relative could not be established, the appropriate city council was asked to provide information regarding whether the patient had died or moved. The period of follow-up was terminated by death or the end of the study period (October 2002 for SE patients; May 2002 for AS patients). This study was approved by the Ethics Committee of the Philipps-University Marburg.
Primary comparisons were performed between patients with SE and patients with AS. The distributions of patients' characteristics were assessed with Fisher's exact test or its extension to RxC tables for categoric variables and Wilcoxon's two-sample test for continuous variables. Survival time was calculated from the date of diagnosis to the date of last contact or death of any cause. Patients lost to follow-up were treated as censored cases at the latest date they were confirmed to be alive. Survival curves were generated according to the Kaplan–Meier (KM) method and compared by the log-rank test. Reverse KM estimators were used to quantify the median follow-up time. Graphical checks based on log–log survival curves and formal tests based on Schoenfeld residuals were applied to assess the proportional hazards (PH) assumption. Because evidence was found that the PH assumption was violated for the group variable (SE vs. AS), an extended Cox model was used to examine the impact of the patient group on the risk of death. Covariates considered potential confounders included age (continuous), sex (female, male), type of stroke (IS, ICH), affected hemisphere (unilateral, bilateral, unknown), and localization of lesions (supratentorial, infratentorial, both, unknown). All tests of significance were performed at the 5% two-sided significance level. Hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) were calculated and considered statistically significant if the CI exceeded 1.0. No adjustment for multiple testing was performed. Thus the results are not confirmative. Statistical analyses were carried out by using SAS software version 8.2 (SAS Institute, Inc., Cary, NC, U.S.A.).
In total, 166 patients met the inclusion criteria and were available for analysis, 93 patients with status epilepticus of vascular origin (SE group) and 73 patients with acute stroke (AS group). Patients' characteristics by group, including patient age at diagnosis, sex, type of stroke, affected hemisphere, and localization of lesions are summarized in Table 1. Forty of the 93 patients in the SE group (43%) had epilepsy before the first episode of SE. No good evidence was seen for differences between the two groups for sex and age, except perhaps for age 85 years and older (p = 0.0469). Patients with AS were more likely to have IS (p = 0.0018), unilateral affected hemisphere (p = 0.0359), and infratentorial lesions (p = 0.0247) than were patients with SE. In contrast, patients in the SE group were more likely to have ICH (p = 0.0018) and lesions of unknown localization (p = 0.0013) than were patients in the AS group.
Table 1. Patients' characteristics
p Values from Fisher's exact test or its extension to RxC tables for categoric variables and Wilcoxon's two-sample test for continuous variables. Squared brackets indicate multiple p values; one for each patient-characteristic stratum. Figures in parentheses are percentages among total number of patients in each group.
SE, status epilepticus; AS, acute stroke; IS, ischemic stroke; ICH, intracerebral hemorrhage.
No. of patients
Type of stroke
A total of 88 (53%) patients died during the study: 53 in the SE group and 35 in the AS group, representing an event rate of 57% and 48%, respectively. There were 25 (47%) deaths of cardiovascular causes in the SE group and 13 (37%) in the AS group. In contrast, deaths of other causes were rather infrequent: three (6%) were observed among SE patients and five (14%) among AS patients. The etiology of the deaths was unknown in almost half of the patients (AS group, 49%; SE group, 47%). Recurrent SE was not documented as the cause of death in any patient. One patient from the AS group and five patients from the SE group were lost to follow-up (four of them without the date of the last patient contact). The reverse KM estimate of the median follow-up was 32.2 months for the SE group and 38.8 months for the AS group (p = 0.1122).
Table 2 shows the median survival times and survival rates at 30 days, 6 months, and yearly intervals by groups of patients. Only the affected hemisphere was significantly associated with survival in the unadjusted analyses (p = 0.0360). In contrast, patient group (SE vs. AS) showed no significant impact on survival (p = 0.0832). Survival rates at 3 years were 37% and 52% for SE and AS patients, respectively. The median survival time was 17.0 months (95% CI, 12.0–29.3) for patients with SE, whereas in the AS group, the median was not reached. Plots of the unadjusted KM curves for SE and AS patients are shown in Fig. 1. The two functions cross at ∼6 months. Before 6 months, the KM curve for the SE group is slightly higher than the KM curve for the AS group, indicating that the survival probability for the SE group is higher during this time compared with the survival probability for the AS group. In contrast, after 6 months, the survival probability for the SE group is lower than the survival probability for the AS group. Thus the hazard ratio is not constant over time. Further evidence that the PH assumption was violated for the group variable was obtained from graphic and formal tests. The estimated log–log plots for patient group were nonparallel (data not shown), and the correlation between the Schoenfeld residuals for patient group and the ranking of individual survival times was 0.18 (p = 0.09). Consequently, an extended Cox model was used to assess the impact of patient group on survival. Based on the graphic results, a cutoff point of 6 months was chosen to specify two functions of survival time. Corresponding to this model, the effect of patient group was described by two distinct hazard ratios, one for time <6 months and the other for time ≥6 months. The results of the extended Cox analysis relating the risk of death to the patient group and the five potential confounders are shown in Table 3. The multivariable analysis confirmed a significant decrease in overall survival when both hemispheres were affected. On multivariable analysis, the risk of death also increased significantly with increasing age. As in univariate survival comparisons, no significant association with sex, type of stroke, and localization of lesions was observed in the multivariable analysis. In contrast, patient group was now found to be associated with survival. The results in Table 3 show a nonsignificant hazard ratio of 0.91 (95% CI, 0.50–1.67; p = 0.7675) for the effect of patient group when time precedes 6 months, but a significant hazard ratio of 2.12 (95% CI, 1.04–4.32; p = 0.0392) when time exceeds 6 months. Even though the latter interval is quite wide, it suggests that after 6 months, patients with SE were at about twice the risk of death than were patients with AS.
Table 2. Determinants of outcome: univariate survival comparisons
Cases n (%)
Deaths n (%)
Median survival (mo)
p Values from log-rank test.
SE, status epilepticus; AS, acute stroke; IS, ischemic stroke; ICH, intracerebral hemorrhage.
Group [p = 0.0832]
Age (yr) [p = 0.3049]
Sex [p = 0.9526]
Type of stroke [p = 0.8150]
Affected hemisphere [p = 0.0360]
Localization [p = 0.2093]
Table 3. Determinants of outcome: extended Cox model
[95% confidence interval]
p Values from Wald test.
SE, status epilepticus; AS, acute stroke; IS, ischemic stroke; ICH, intracerebral hemorrhage.
Group × Time
Group × Time
Type of stroke
Finally, several models were explored by treating age as dichotomous variable (<73, ≥73 years, median split), recategorizing localization (supratentorial vs. infratentorial vs. both or unclear), and choosing another time point as the cut-off (170 days), without altering the results.
The present study aimed to determine whether SE is a risk factor for an increased long-term case fatality rate. To exclude etiology as an established predictor of short-term and long-term case fatality (Oxbury and Whitty, 1971; Scholtes et al., 1994; Towne et al., 1994; DeLorenzo et al., 1995; Logroscino et al., 1997; Logroscino et al., 2002), patients with SE of cerebrovascular origin were compared with patients with acute stroke, but without seizures. The main result of this study is that patients with the first episode of SE of vascular etiology have a significantly higher long-term case fatality as compared with patients with an AS without SE. This finding of SE group as an adverse prognostic factor with a hazard ratio of 2.12, even after adjusting for five potential confounders, was strengthened by the multivariable model.
The influence of differences between patient groups
Even though the data of both subpopulations of the study were collected by the same department of neurology, patients were not matched, and significant differences between patient groups must be considered (see Table 1). Median age was slightly younger in the SE group, but this difference was not significant and should otherwise have resulted in a lower long-term case fatality in this group (Logroscino et al., 2002). The presence of a basilary artery syndrome is a predictor for an increased short-term case fatality (Anderson et al., 1994; Sacco et al., 1994). Therefore the higher percentage of infratentorial lesions in the AS group should have resulted in a higher rather than a lower case fatality. However, populations have not been analyzed for the occurrence of the prognostic factors of morbidity of patient with stroke, such as Glasgow Coma Scale (Handschu et al., 2005) or cardiovascular risk factors.
The percentage of unknown localizations was higher in the SE group. This is likely due to the difference in data acquisition.
ICH was more frequent in the SE group, and subsequent ISs were more frequent in the AS group. These differences likely reflect the fact that patients with ICH have a higher risk to develop seizures and SE as compared with those with IS (Labovitz et al., 2001). Patients with ICH have a higher early case fatality than do those with ISs. In CT-based studies, patients with ICH showed short-term case fatality rates between 35% and 55% (Broderick et al., 1989; Bamford et al., 1990; Rosenow et al., 1997). However, at 2-year follow-up, the case fatality increased by only 3.4% to 38% in one study (Rosenow et al., 1997). Conversely, patients with IS have a lower short-term case fatality of ∼10% at 30 days and 20–25% at 1 year, depending on stroke type, extension, and resulting disability (Anderson et al., 1994; Sacco et al., 1994). However, long-term case fatality is ∼50% and thus not significantly different from that of patients with ICH (Dennis et al., 1993). Similarly, in this study, the long-term case fatality was not higher in patients with ICH as compared with those with IS. ICH was not an independent predictor of increased case fatality.
In the SE group, patients with remote stroke were included. It was previously reported that patients with SE caused by a remote stroke have a lower case fatality than do patients with an AS and no SE (Waterhouse et al., 1998). This is the likely cause of the observed lower short-term case fatality as compared with the AS group during the initial 6 months. This should have also resulted in a decrease rather than an increase in long-term case fatality in the SE group and does not explain the observed difference in the long term (Waterhouse et al., 1998). The increased long-term case fatality of the SE group is, therefore, not a consequence of differences in short-term survival, but has to be attributed to other factors. Rumbach et al. (2000) reported a case fatality rate of 50% in 30 patients with stroke-related SE who were followed up for a mean of 47 months. Only one of the 15 patients died within the first 30 days, resulting in a long-term case fatality rate of 47%. Similarly, in the present SE group, the estimated probability of death at 3 years for those who survived the first 30 days was 57%.
No evidence indicated that poorer prognosis of the SE patients could have been due to death in recurrent SE. However, as the cause of death was unknown in almost half of the patients, it cannot be ruled out that recurrent SE contributed to the increased case fatality. In a previous study, death was related to SE in five of 15 patients and of cardiovascular origin in another third of the patients (Rumbach et al., 2002).
A 5-year risk of death for patients with IS is age dependent and lies between 40% and 60% (Dennis et al., 1993). The estimated long-term probability of death in the AS group of this study stayed within this range (48% at 3 years), whereas the SE group exceeded this range (63% at 3 years). Therefore the significant differences were unlikely to be due to the selection of the unmatched control group.
Factors contributing to the increased case fatality after SE
Established predictors of long-term case fatality after IS are age, cardiac disease, and severity of the neurologic deficit. The main cause of death is vascular disease (Sacco et al., 1982; Howard et al., 1991). Patients die of cardiovascular disease 2 times more often than from a stroke (including index stroke and stroke recurrence) (Sacco et al., 1994). Similarly, 43% of deaths were cardiovascular in this study (i.e., 83% of ascertained causes of death), suggesting that this was the most frequent cause of death.
Bilateral as compared with unilateral cerebrovascular disease was associated with a hazard ratio of 2.28 in the multivariable analysis. Bilateral disease, conversely, is a marker for severe arteriosclerosis and/or cardiac disease (Roh et al., 2000; Kang et al., 2003) with a high risk of cardiovascular death (Kaarisalo et al., 1997; Wong and Li, 2003; Stollberger et al., 2004). Conversely, bilateral cerebrovascular disease represents more extensive disease that is associated with an increased risk to develop seizures or SE (Arboix et al., 1997; Labovitz et al., 2001). Therefore one hypothesis contributing to the increased case fatality in the SE group observed is that SE in patients with cerebrovascular lesions reflects extensive cerebrovascular disease and indicates an increased risk of cardiovascular death, which remains high in the long term. In contrast, Waterhouse et al. found a striking increase in short-term case fatality in patients with AS and SE as compared with those with only AS (Waterhouse et al., 1998). Because this difference was not explained by differences in age or stroke size (severity), they concluded that the occurrence of SE in this setting directly contributed to a higher short-term fatality.
Shortcomings of the study and impact on future research
The main limitations of this study include limitations inherent in any retrospective approach. Patient selection and data ascertainment did not allow to investigate systematically the influence of all aspects of stroke etiology, extension and localization of cerebrovascular lesions, and neurologic deficit on case fatality. A sequential stroke registry–based control group was used resulting in significant group differences regarding several factors, which were, however, included in the multivariable analysis. Therefore the results of this study are of clinical interest and relevance. Certainly, further well-designed, prospective studies are necessary to clarify the validity of the association found between SE and increased long-term case fatality. Data regarding concomitant diseases, especially atrial fibrillation, severe cardiac disease, and arteriosclerosis should to be ascertained systematically. In future studies, it would also be favorable if patient groups were matched regarding lesion size, localization of lesions, and type of stroke.
Patients with SE of cerebrovascular etiology should be considered a population at a high risk of death within 3 years. In contrast, the case fatality risk attributable to recurrent status or seizures is lower. Therefore patients need to be closely followed and appropriate secondary preventive measures should be applied in respect to cardio- and cerebrovascular disease in addition to AED.
Acknowledgment: We thank the Verein zur Erforschung der Epidemiologie der Epilepsien e. V., Germany and Drs. I. Sünkeler and H. Pape for their support and contribution.