The most frequently studied clinical outcome is mortality. Several different observational cohorts, population-based studies, or hospital discharge databases provide information about fatality rates following SE. Most focused on short-term mortality, e.g., at the time of hospital discharge, or at 1 month. Due to study design and inclusion criteria, however, the fatality rates are remarkably varied, ranging from 3–9% [21, 22] to 19–39% [23, 24]. The four studies cited were very different, the first two derived from nationwide hospital discharge databases, and the other two prospectively collected data from a single center and in a population-based assessment, respectively. Another very important factor is whether studies include children or not; mortality is usually lower in children than in adults, probably due to differences in etiologies and, to a lesser extent, comorbidities [25]. For example, a population-based assessment performed in French-speaking Switzerland that included children reported a mortality rate of 8% [1], while rates of 9–37% were found in two studies with similar design but including adults alone [24, 26]. It is also important to note that post-anoxic myoclonic SE has been included in some analyses––which may dramatically increase mortality rates, because most patients with this condition ultimately die [27]. As an example, the study by DeLorenzo et al. [2] which included patients with anoxic brain injury, reported a mortality rate of 22%, while a recent multicenter prospective cohort excluding post-anoxic cases reported a rate of 12% [28]. Table 9.1 provides information about mortality rates across different studies. An overall 10–15% short-term fatality rate appears reasonable to expect following GCSE in adults.

Few studies address the long-term effect of SE on mortality. A population-based survey in Rochester, Minnesota reported that more than 40% of patients who survived for the first 30 days after SE had died by the time of 10-year follow-up [29], representing a threefold increase in mortality as compared to that in a matched population without SE. All seizure types were included, 47% with GCSE. Another study including mostly GCSE (76.4%), reported a mortality rate of 20% at 12 years [30]; but follow-up data were available for only 32% of SE survivors, and the proportion of GCSE in follow up was not reported, suggesting that this might be a substantial underestimation of mortality.

Mortality rates after SE may depend significantly on demographic features, on the different types and causes of SE, and on different approaches to treatment.

The Patient: Some demographic factors such as gender appear to have no clear influence on mortality [31], but race may have a role: one study found an higher frequency of SE in non-whites as opposed to Caucasians (71 vs. 23 per 100,000 people, respectively) across all age categories, but especially in the very young and in the elderly. Mortality, however, was higher in Caucasians (31% vs. 17%) [2]. In several studies, increased age has been associated with mortality, with fatality rates increasing from 1% in the age group 1–19 years old, up to 47% in older groups (>65 years) [29]; being over 65 years old was associated with a 5.4 odds ratio (OR) for mortality after correction for most known confounding factors [32]. Of note, increased age is not only associated with higher mortality but also with a higher incidence of SE [2, 25], probably due to the increased burden of structural brain lesions (strokes, tumors, etc.).

Furthermore, medical comorbidities increase with age, but there are only a few studies addressing this topic. One US hospital discharge database indicated that patients with a greater number of comorbid conditions have a worse outcome [21], but this study had several limitations, including retrospective design and use of coded diagnoses. A prospective European study found that comorbidities, quantified using the Charlson Comorbidity Index [33], had only a marginal impact on SE mortality when others important factors were accounted for [34].

The Seizures. Seizure semiology is an important factor. While in absence SE the response to anti-seizure drugs is usually prompt and the outcome usually excellent [35], a mortality of up to 67% has been reported in nonconvulsive SE in coma [36]. GCSE lies somewhere in between. There was a mortality rate of 27% in a US population-based study of adult patients [2]; 21% in a retrospective hospital cohort from Turkey [37]; 10.1% in a subgroup analysis of a prospective hospital cohort of patients in Massachusetts and Switzerland [5]; 9% in a prospective cohort in western France [38]; and as low as 3.5% in the US hospital discharge database [21]. Even within a single seizure or SE type, mortality may vary remarkably. Based on these numbers, a reasonable estimate of mortality after GCSE is around 10–15%.

Analysis of the impact of SE duration has generated conflicting results. One study found a strong association between worsened outcome and a threshold of 1 h SE duration [39], while another study showed that while survival was better if SE lasted less than 10 h, SE duration lost its predictive value for poorer outcome once etiology, presentation in coma, and type of SE were accounted for [40]. This suggests that SE duration exerts probably a minor independent prognostic role.

Refractory SE (RSE), defined as ongoing clinical or electrical seizures despite two adequate lines of treatment (a benzodiazepine followed by a non-sedative anti-seizure drug) [4], confers a high rate of mortality; studies focusing on RSE report mortality rates ranging from 16.7 to 39% [41–43] (Table 9.2). Super-RSE (SRSE), corresponding to persistence of seizures even after 24 h of coma induction [44], seems to herald an even worse outcome. Unfortunately, and probably due to the relatively low incidence of SRSE, comprehensive data are sparse, relying mostly on case series and small cohorts not limited to GCSE [45–49]. As shown in Table 9.3, however, a relatively favorable outcome is possible even after very prolonged SE durations, but mortality rates are still around 30%, with most survivors suffering from new morbidity.

The role of level of consciousness has been well established in several studies [32, 36, 40, 42]. For example, an increase of mortality with an OR of 3.03 has been reported for each decrease in the level of consciousness, from aroused, confused, to stuporous or comatose [32].

Electroencephalography (EEG). Some studies indicate that EEG background organization and sleep architecture correlate with preserved cognition and brain metabolism [50], suggesting that EEG might provide clues for outcome prediction. There are conflicting results in the literature. Ictal discharges occurring after SE, burst-suppression patterns and, to a lesser extent, lateralized periodic discharges, have been associated with higher mortality [51]; but in that study, 20% of patients had post-anoxic SE. Another retrospective analysis that did not control for confounding factors found an increased mortality in patients with lateralized periodic discharges [52]. More recently, a prospective assessment using validated EEG terminology found that EEG background (including posterior dominant rhythm and sleep architecture) were the only reliable outcome predictors after correction for SE severity and etiology [53]. Of note, lateralized periodic discharges were associated with an increased mortality in patients without brain injury in a case-control study of patients undergoing EEG monitoring [54].

The Etiology. Many demographic and clinical aspects are major determinants of mortality after GCSE, but the most important one is probably the underlying etiology itself [55]. In two epidemiologic studies, mortality related to SE due to low levels of anti-seizure drugs in patients with epilepsy was 4% [2] and 2% [25], as opposed to 33% with SE and central nervous system (CNS) infections, or 25% with stroke [25]. To address its independent role in SE outcome prediction, etiology has often been categorized as acute or not [2, 23, 32, 56], with some studies reporting an increased mortality for patients with an acute etiology, with ORs ranging from 0.4 [23] to 2.2 [56]. Of note, loss of predictive value after adjustment for other factors has also been reported [32]. This is probably due to the marked heterogeneity of “acute” etiologies, which include medication non-adherence and alcohol withdrawal, but also acute brain injuries such as stroke, hemorrhage, acute brain infections, and head trauma. To overcome this concern, categorization according to “potentially fatal” etiologies (including: acute large vessel ischemic stroke, acute cerebral hemorrhage, acute central nervous system infection, severe systemic infection, malignant brain tumor, AIDS with CNS complications, chronic renal insufficiency requiring dialysis, systemic vasculitis, metabolic disturbance or acute intoxication sufficient to cause coma in the absence of SE, eclampsia, and intracranial tumor surgery) has been proposed [32]. This approach produced a more robust predictor of mortality, with an OR of 11.7 [32], 6.9 [5], or 5 [57] in different studies.

Because etiology is such an important outcome predictor and because 42% of SE are related to an underlying etiology that requires tailored treatment (in addition to anti-seizure drugs) [28], identification of the etiology of SE should be one of the main objectives of SE management in the acute setting.

The Treatment. There is evidence that treatment guidelines are relatively poorly followed in daily clinical practice [5, 38, 58, 59]. While adequate first-line treatment is strongly associated with interruption of SE [38], its impact on mortality remains unclear. One study found that “insufficient treatment” was more frequent in the non-survivor group than in survivors (45% vs. 22% respectively) [60]. Better medical management was strongly and independently related to better clinical outcome (OR 21.09) in an Italian study comparing a peripheral and a university hospital [6]. Conversely, a Canadian cohort reported a comparable outcome between patients with or without “appropriate” treatment [61]. Finally, adherence to treatment guidelines, assessed by deviations from drug treatment sequence or dosages (±30% of the recommended doses) did not appear to have a significant impact on mortality [55]. While to our knowledge, this was the only study to include all known predictors in statistical models (demographics, etiology, SE severity, comorbidity, and treatment adequacy), it should be acknowledged that the definition of “treatment deviation” is critical and varies significantly from one study to the next.

Therapeutic coma using anesthetic drugs is recommended when second-line anti-seizure drugs fail to control seizures [4], but the evidence supporting this practice is sparse [62] and there are potentially serious side effects [63]. Some observational cohorts have questioned this practice. One was a retrospective assessment of 126 SE patients treated in ICUs in Baltimore [64]. The use of anesthetic drugs was associated with an increase in mortality (OR: 8.65), after adjustment for age, de novo SE, and etiology. Another study, of a prospective cohort of 171 patients with SE treated in the ICU in Basel, Switzerland, found a 2.9-fold relative risk of death when anesthetic drugs were used [65]. Of note, SE duration, severity of SE, use of further non-anesthetic antiseizure drugs, critical medical conditions, and etiology were considered in the analysis. Therapeutic coma was associated with poorer outcome (relative risk ratio for mortality, 9.10) in a prospective cohort of 467 patients in Lausanne, Switzerland [66]; the effect was greater in patients with focal as compared to GCSE, or nonconvulsive SE in coma. Demographics, etiology, impairment of consciousness, SE severity, treatment latency, and comorbidities were accounted for. Another study concentrating on de novo refractory SE, however, did not confirm these results [67], even after adjustment for SE severity, burden of complications, and SE duration. It remains difficult currently to distinguish between confounding by the treatment indication and the negative effects of intubation and coma induction. Nevertheless, a prospective controlled study would be very difficult to plan and carry out in this setting.

Unfortunately, while all studies report mortality rates after SE, information regarding reasons for death is missing. One study attempted to address these details [68], based on 920 SE patients managed in a single center over 10 years. The vast majority (78.8%) had GCSE. Of 120 deaths, 65.8% were attributed to underlying illnesses alone; the underlying etiology was not involved in the explanation of death in only 14 cases. In that subgroup, death was related to coma or treatment complications.

To date, there are only two studies with well-documented cognitive measurements obtained both before and after SE. Both describe cohorts of patients living with epilepsy. In the first [69], only mild intellectual consequences were reported in nine adult patients with SE (four in GCSE) in comparison to nine control patients without SE. The second study [70] failed to demonstrate any changes in cognitive measurements from before to after SE in 15 patients, nor in 40 matched patients with epilepsy, but without SE, during the same period. Of note, most patients had secondarily generalized tonic–clonic seizures and a median duration of SE of 240 min.

Some studies assess SE functional outcome using more rudimentary scales, such as the Glasgow Outcome Scale (GOS). A multicenter prospective cohort study from France with 177 GCSE patients, found that 18.8% were dead at 3 months follow-up; half of the survivors had severe functional impairment; 42.5% had a good outcome, with a GOS of 5 (i.e., the patient was able to return to work or to school). Longer seizure duration, cerebral injury, and RSE were strongly associated with poorer outcomes [71]. A single center retrospective study assessed 83 SE episodes and found that of 69 survivors, 16 (23%) had a deterioration of at least one point on the GOS at discharge compared to that on admission. Factors associated with functional deterioration after multiple logistic regression were acute symptomatic seizures and length of hospitalization [72]. Finally, one study used the modified Rankin Scale (mRS) to evaluate functional outcome after RSE: out of 63 SE episodes, poor functional outcome (mRS 4–6) occurred in 76.2%, and only 8 (12%) regained their premorbid status [49]. Of note, a few patients improved several months after the acute facility discharge. Duration of drug-induced coma, cardiac arrhythmia requiring intervention, and lung infection were associated with poor functional outcome.

Another outcome of interest after de novo SE is the risk of developing epilepsy: SE appears to increase the odds for subsequent unprovoked seizures 3.3-fold as compared with initial brief acute symptomatic seizures, after controlling for age, gender, and cause [73], but a specific role for the generalized convulsive form of SE remained uncertain because the majority (41%) had a SE with focal seizures and only 23% of patients had GCSE. More recently, in a prospective cohort of 89 patients with de novo SE, 58.7% had recurrent seizures after a median follow-up of 10 months [74]. After multivariable analysis, development of subsequent epilepsy was associated with SE lasting more than 24 h only, with an OR of 3.8. This is in line with a previous finding that secondary epilepsy was more frequent after RSE than non-refractory SE [41].

Finally, a functional outcome assessed in some studies is the “return to premorbid baseline”, meaning that the patient’s functional status at hospital discharge is the same as before the SE episode. This outcome measurement, however, is somewhat subjective and maybe imprecise, because complete recovery is often difficult to prove. Nevertheless, among 27 patients with SE due to alcohol abuse, most patients (81.5%) returned to their clinical baselines at discharge [75]. It was also pointed out that a potentially fatal etiology (see above) was less common in patients returning to their premorbid conditions (34.3%) than in patients who did not return to baseline (72.1%) [32]. These two studies emphasize, once again, the cardinal role of etiology in the likelihood of returning to the patent’s premorbid baseline. Our group also showed that increased age, more severe SE, and an increased load of comorbidities, decreased the probability of regaining a patient’s clinical baseline, and conversely, that neither the choice of second-line anti-seizure therapy nor adherence to treatment guidelines influenced this outcome [55, 76]. The EEG may also provide some information: the finding of normal stage 2 non-REM sleep patterns (such as K complexes and spindles) is associated with an odds ratio of 2.6 favoring a complete recovery [53].