- Top of page
Purpose: Outcome for patients with status epilepticus (SE) depends strongly on etiology. Duration of SE is also predictive, at least in the first 2 h, but beyond this it is unclear that duration of SE influences outcome significantly. We sought to determine the influence of duration of SE on outcome in patients with prolonged SE, and to compare this influence with that of other factors.
Methods: We reviewed the clinical course and outcome of 119 patients with SE, diagnosed by both clinical manifestations and electroencephalography (EEG) evidence. Using univariate and multivariate analyses, we sought predictors of outcome (survival vs. death or vegetative state) among age, etiology (epilepsy, anoxia or severe hypoxia, or other), presence of earlier epilepsy, multiple medical problems, presentation in coma, and type of SE (focal or generalized).
Results: Median duration of SE was 48 h. Survival was greater with a shorter duration, especially when <10 h (69% vs. 31% for longer duration; p < 0.05). Epilepsy as the etiology, and an earlier diagnosis of epilepsy offered a favorable prognosis (p < 0.01), but only the former on multivariate analysis. Coma and SE caused by anoxia/hypoxia were unfavorable factors. Once corrected for etiology, presentation in coma, and type of SE (focal or generalized), duration of SE did not have a significant effect on outcome. Overall mortality was high, 65%, but 10 patients survived SE lasting over 3.5 days.
Conclusions: A duration of <10 h was associated with better outcome in SE, but this was not significant once etiology, presentation in coma, and type of SE were accounted for. Etiology of SE is still the primary determinant of outcome. Unless it follows anoxia, prolonged SE should not be considered a hopeless condition.
After etiology, duration has often been considered the most useful factor in predicting outcome of status epilepticus (SE) (Aminoff & Simon, 1980; Towne et al., 1994). Although etiology is already determined at the presentation of SE, duration is considered a modifiable factor, amenable to treatment and thus an improved outcome. Patients with prolonged SE have a high risk of mortality (Towne et al., 1994; Claassen et al., 2002a; Holtkamp et al., 2005), and it is sometimes assumed that beyond some point, prolonged SE becomes very difficult to treat, or even hopeless. Nevertheless, there are also reports of patients with very prolonged SE (lasting weeks to months) with good outcomes (Mirski et al., 1995).
The relatively few studies showing a worsened prognosis with prolonged SE have concentrated on the first few hours (Lowenstein & Alldredge, 1993; Scholtes et al., 1994). The best available study on the prognostic significance of SE duration described a 2.7% mortality for episodes lasting less than an hour, but 32% for those lasting more than an hour (Towne et al., 1994). That study included control for different etiologies, showing that duration could be established as an independent predictor. It was not clear, however, that any duration greater than 2 h, or even beyond 24 h, led to a statistically significant further increase in mortality. That report also confirmed a major influence of etiology on outcome, especially for anoxia (with extremely poor outcomes) and either alcohol-related factors or discontinued anticonvulsants (with better outcomes). It left open the possibility that patients with very prolonged SE were also the patients who had more ominous etiologies for SE, or those who were not treated adequately.
In order to explore further the role of duration on outcome in prolonged SE, we reviewed 119 cases of SE determined by both clinical and electroencephalography (EEG) evidence of SE. To determine if duration was an independent predictor of outcome, we also examined the influence of etiologies (worsened epilepsy, anoxia/severe hypoxia, or others) as well as those of level of consciousness, an earlier diagnosis of epilepsy, the presence of multiple medical problems, and type of SE (focal or generalized).
- Top of page
Patients were identified from EEG records from the preceding 20 years at Beth Israel Deaconess Medical Center (BIDMC), and medical records were reviewed for clinical details after approval by the BIDMC Institutional Review Board (IRB). EEG studies were read by two electroencephalographers certified by the American Board of Clinical Neurophysiology (FWD, DLS) as showing ongoing seizures with either brief, intermittent evolving electrographic seizures, or prolonged but rhythmic and rapid epileptiform discharges, either focal or generalized, but not including periodic lateralized epileptiform discharges with discharge intervals longer than 1 s. Clinical records confirmed an abnormal mental status or focal neurologic deficit lasting at least 30 min, including at the time of the EEG recording. Treatment of SE was not carried out according to a standardized protocol but was the province of many different attending physicians.
Outcome (at the time of discharge), duration of SE, age, etiology, level of consciousness at the time of diagnosis, an earlier diagnosis of epilepsy (whether or not the cause of this episode of SE), and the presence of multiple medical problems were ascertained. Patients with focal motor abnormalities on clinical examination or with predominantly focal EEG findings during SE were considered in focal (as opposed to generalized) SE. No patients had simple partial SE, or primary generalized epilepsies. Outcome was considered good if the patient survived, even if there were persistent neurologic deficits, for example, caused by new strokes. Poor outcomes included death or a vegetative condition.
Duration of SE could not be determined exactly because very few patients had continuous EEG throughout the illness. The onset of SE was ascertained at the beginning of clear clinical seizure activity or from a sudden and major decline in neurologic function (without interval recovery) before the first EEG showing ongoing seizures. Cessation of SE was determined by clinical recovery, if this occurred rapidly, or from EEG evidence. SE was considered to have ended with the last EEG showing ongoing seizures when the patient recovered gradually thereafter or if the patient died at any point after that EEG. The possibility could not be excluded that SE continued beyond the final EEG showing ongoing seizures or that, in some patients, SE may have ceased between intermittently recorded EEGs and resumed before the next EEG.
Association between outcome and potential predictors was first assessed by univariate logistic regression. Then, factors with a univariate significance of <0.15 were entered into a multivariate logistic regression model, using the SAS “Logistic” procedure (SAS corporation, Cary, NC, U.S.A.). The effect of an independent variable on the outcome in the multivariate setting was assessed by repeated process of inclusion and exclusion of that variable in the model. Variables with a significance level of <0.05 were presented in the final model. Chi-square and Fisher’s exact tests were also used for categorical data. In situations where a cell (in a 2 × 2 table showing outcome and predictor) did not have any observation, a small number was added to the cell, or the logistic regression was run with a small weight for that cell, or exact logistic regression was used. For purposes of illustration, patients were also divided into eight groups of approximately equal numbers according to duration of SE. Following this, the logistic regression was carried out again, just for patients with duration >10 h.
- Top of page
In all, 119 adult patients were identified with both clinical and EEG evidence of SE. Age ranged from 24–96 years (mean 67 years). Durations of SE ranged from 45 min to 20 days, with a median of 48 h; all but one were >1 h.
Outcome was poor in this series—35% survival overall, 44% without anoxia. Significant predictors of outcome by univariate analysis included the etiologies of epilepsy or anoxia, duration of SE, coma upon presentation, an earlier diagnosis of epilepsy, and generalized forms of SE (Table 1). Age and multiple medical problems were not predictive. Upon multivariate analysis, SE caused by epilepsy, and presentation in coma were predictive, but duration of SE and an earlier diagnosis of epilepsy were not (Table 2). Anoxia was strongly predictive of a poor outcome, but anoxia was highly correlated with presentation in coma; all anoxic patients were comatose (and all had generalized epileptiform discharges). There were many more comatose than anoxic patients, and coma was a better predictor in multivariate analysis. Generalized (as opposed to focal) SE also remained a significant predictor of a poor outcome.
Table 1. Prognostic factors for outcome in prolonged status epilepticus univariate analysis
|Factor||OR (for poor outcome)|| 95% CI||p-Value|
|Etiology: anoxia/hypoxia ||19.0||2.47–146||<0.005|
|Earlier diagnosis of epilepsy||0.223||0.096–0.523||0.005|
|Coma on presentation||33.3||11.1–100||<0.0001|
|Generalized epilepsy/discharges ||2.24||1.10–5.32||0.0273|
|Duration in hours (continuous)||1.007a||1.001–1.013||0.0243|
|Multiple medical problems||1.90||0.89–4.06||0.16 (N.S.)|
|Duration > 10 hb||4.30||1.21–15.3||0.0242|
|Duration > 24 hb||1.8||0.74–4.36||0.19 (N.S.)|
Table 2. Prognostic factors for outcome in prolonged status epilepticus multivariate analysis
|Factor||OR (for poor outcome)|| 95% CI||p-Value|
|Coma on presentation||152||18.6–>1,000||<0.001|
For purposes of illustration, patients were divided into eight approximately equal-sized groups by duration, with survival shown in Fig. 1. The only group with a survival statistically different from that of all later groups was that of patients with SE durations less than 10 h (69% survival vs. 31% for all others; p = 0.024) (Table 1). Repeating the multivariate analysis, this time using duration of greater than or less than 10 h as one factor, showed that duration was no longer predictive, whereas SE caused by epilepsy, presentation in coma, and generalized SE still were.
Ten of 39 patients with SE durations ≥3.5 days survived (26%). Of these, four had epilepsy as the cause (although two of these were comatose on presentation) and the others had many different causes, including infection and underlying malignancy (but none of these without epilepsy presented in coma). Two of the seven patients with neither epilepsy nor anoxia whose SE lasted over 8 days survived.
The 33 patients (28% of the total) with earlier diagnoses of epilepsy had somewhat shorter SE durations and included 7 of the 12 patients with durations <10 h. Of patients with earlier epilepsy, 18 (55%) survived, versus 24 of 86 (28%) without epilepsy (p < 0.01). There was still significant mortality in this group. Two epilepsy patients with SE durations of 5–7 days survived, but the other four with durations of ≥3.5 days did not.
Patients with SE caused by anoxia or severe hypoxia were overrepresented among those with longer SE duration (median 72 h): just 3 of 24 patients had durations shorter than 48 h, and none under 12 h. All 24 of these patients died—markedly different from the 44% survival of patients without anoxia.
Focal SE (vs. generalized) was a predictor of a better outcome in both univariate and multivariate analysis. Eight of 13 patients age 40 years or younger survived, including all five epilepsy patients younger than 40, but age was not a statistically reliable predictor.
- Top of page
In these patients, most of whom had prolonged and refractory SE (Jagoda & Riggio, 1993; Mayer et al., 2002), outcome worsened with longer duration of SE. In a secondary analysis (after viewing Fig. 1), an SE duration less than 10 h predicted a better outcome. Nevertheless, neither overall duration (assessed continuously) nor duration greater than 10 h was a statistically reliable predictor upon multivariate analysis including etiologies, presentation in coma, and the type of SE—focal or generalized.
The prognostic effect of duration is influenced markedly by etiology. Patients with anoxia/hypoxia did extremely poorly, whereas those with epilepsy as the cause of SE did much better. The apparent beneficial effect of shorter duration on outcome may derive, in part, from the fact that patients with epilepsy tend to have shorter durations of SE and those with anoxia and other severe illnesses, longer durations. Much of the effect of duration on outcome in this and other studies appears dependent on etiology, still the most important prognostic factor.
One earlier report also found that duration was not a significant risk factor for short-term (30 days) mortality in SE once adjusting for etiology (Logroscino et al., 1997). An SE duration of less than1 h offers a favorable prognosis (Towne et al., 1994), but SE durations may be extended early in its course by delayed diagnosis, especially in critically ill patients (Drislane et al., 2008) or by inadequate treatment (Cascino et al., 2001). Later, duration may be extended by the frequent unwillingness of treating physicians to give up on patients with SE caused by anoxia or other etiologies with poor prognoses. After the first few hours, duration of SE is a less reliable predictor of outcome.
Other prognostic factors are familiar. Coma is a poor sign (Rossetti et al., 2006), and similar comatose patients with SE had a mortality greater than 50% in another series (Lowenstein & Aminoff, 1992). In nonconvulsive SE, coma also augers poorly (Shneker & Fountain, 2003). Presentation in coma, however, may also reflect the severity of the underlying illness. A larger study of determinants of mortality in SE found that age younger than 60 predicted a better outcome (Towne et al., 1994). Age younger than 40 appeared favorable here, but this did not reach statistical significance, possibly because there were fewer younger patients evaluated. Finally, focal SE predicted a better outcome here, possibly deriving from an etiology of non–life-threatening stroke. This was not the case in the series noted earlier (Towne et al., 1994), possibly because of a different patient mix, in turn related to the use of EEG and the selection of sicker patients with generalized discharges and generalized SE in this series.
Mortality was very high in this series (56%, even eliminating patients with anoxia), but this is not unusual for patients with prolonged or refractory SE, especially when there is an acute symptomatic cause (Barry & Hauser, 1993; Delanty et al., 2001), as was the case in most of these patients. In the largest prospective trial of antiepileptic drug (AED) use for SE, survival was only 35% among patients with “subtle” SE (Treiman et al., 1998). Other studies of refractory SE report a mortality of 30–50% (Sahin et al., 2001; Claassen et al., 2002a).
Patients in this study were selected according to both clinical and EEG criteria, thus constituting a relatively ill and refractory group. Patients with SE who respond quickly to AEDs do not have or need EEG confirmation. In a population-based study of patients with all forms of SE, mortality was about 22% overall and 26% in adults (DeLorenzo et al., 1996). Therefore, the course of patients with particularly prolonged SE may not be representative of all patients with SE.
Particularly prolonged cases of SE are sometimes referred to clinically as “burnt out,” even when convulsions end, and are often thought to have a very poor likelihood of survival. Although a longer duration cannot be salubrious, a primary lesson from this series should be that patients with prolonged, refractory SE should not be considered hopeless solely on the basis of long SE duration. Good outcomes are still possible—not significantly different in this series for SE lasting 3.5–7 days from that lasting 12–24 h (Fig. 1). One-third of the survivors in this series had episodes estimated to last more than 50 h, and there are several reports of patients with good outcomes after SE lasting weeks or even months (Mirski et al., 1995; Krishnamurthy & Drislane, 1997).
Limitations of this study include the possibility of a beta error, that is, a much larger, similar, population might be sufficient to demonstrate a poorer prognosis with longer SE duration. Similarly, it did not include enough patients to determine the role of duration among finer etiologic categories such as stroke, encephalitis, AED withdrawal, among others. In addition, there was no treatment protocol. Furthermore, the study was retrospective, and duration of SE is not a factor subject to randomization.
Finally, without continuous EEG monitoring, the duration of SE could not be established precisely in all patients. Because the termination of SE was fixed at the last EEG showing it, durations are probably underestimated. It is also possible that some patients went from SE to a postictal state and then back to SE again, but with the usual definition of SE as failure to recover between seizures (Commission on Epidemiology and Prognosis, International League Against Epilepsy, 1993), they would still be classified reasonably as in SE for the entire period. It is also likely that SE estimated as lasting 5 days was longer than that estimated at 4 days, and although exact durations may be impossible to determine, the general trend appears likely to remain the same with 119 patients.
It is important to state what this paper does not indicate. It does not provide evidence to support a less-urgent approach to SE early in its course. Seizures lasting 10–29 min often stop on their own, but very few seizures lasting over 30 min do so (DeLorenzo et al., 1999); mortality is substantially higher in the latter group. In another series, 80% of patients with seizures lasting under 30 min responded successfully to AEDs, compared to 40% with SE for more than 2 h, although this also depends somewhat on etiology (Lowenstein & Alldredge, 1993). In another study of patients with generalized convulsive SE, morbidity and mortality attributed to the SE itself (rather than to an underlying illness) increased substantially when duration of SE exceeded 4 h, but not necessarily progressively after that (Scholtes et al., 1994).
In the large, Richmond, Virginia series, SE duration of more than an hour was associated with a 32% mortality, versus 2.7% for that less than an hour (Towne et al., 1994). This was by multivariate analysis, including control for etiology. Still, there was no definite statistically significant decline in survival among patients with durations of 1–2, 2–4, 5–10, 11–23, or more than 24 h.
Physiologically, there appear to be deleterious effects of SE that worsen after about 30 min in humans, at least for generalized convulsive SE and possibly during nonconvulsive SE (Lothman, 1990). Experimentally, seizures become more refractory to treatment with benzodiazepines after 45 min, compared to seizures treated after 10 min (Kapur & Macdonald, 1997). With time, SE becomes more refractory (Chen & Wasterlain, 2006).
For all these reasons, SE should be treated expeditiously. There is a strong clinical rationale for diagnosing and treating SE within 5 or 10 min, rather than waiting for a “classical” definition of a 30-min duration (Lowenstein et al., 1999). In addition, duration of SE is considered changeable by medical intervention. We are unaware of any study showing an improved survival with longer duration of SE.
Longer duration of SE is unlikely to be a positive prognostic factor, but proving that it is a significant negative prognostic factor independent of etiology is very difficult. This study suggests strongly that etiology is still, by far, the most important prognostic factor: Patients with exacerbations of earlier epilepsy and patients with focal onset of SE do better, and those with anoxia almost always die. Level of consciousness (comatose or not) and type of seizure (focal or generalized) are correlated with outcome but probably related to etiology as well. After the first several hours, there is no substantial drop-off in survival with any particular duration of SE and, once etiology is accounted for, there is no major effect of duration on outcome in SE. When the etiology is considered at least permissive of a reasonable recovery, some patients with particularly long durations of SE can still be treated successfully.