Overall mortality in our cohort (38%) was similar to some prior reports, which describe a mortality ranging from 39% to 65% (Cooper et al., 2009; Drislane et al., 2009; Novy et al., 2010), but higher than in other studies (8–26%) (Mayer et al., 2002; Rossetti et al., 20044; Holtkamp et al., 2005; Rossetti et al., 2005), possibly resulting from the high proportion of patients with HE. The high mortality in our cohort may also reflect the retrospective and ICU-based nature of this study. Relative risks of death in patients with brain tumors and HE were 2–3 times higher than for patients with other etiologies in the multivariable analyses, indicating that they are independently associated with mortality. Although the adverse effect of HE after cardiac arrest on outcome in patients with RSE is well described, relative risks were unexpectedly high and have not been reported in such a large sample size. HE after cardiac arrest is known for having a substantial and deleterious influence on mortality (Young et al., 1990; Wijdicks et al., 1994; Hui et al., 2005; Thomke et al., 2005; Rossetti et al., 2007; Kawai et al., 2011; Rittenberger et al., 2012). Remarkably, a high proportion (8/25) of patients with HE survived—an unexpected result that might be at least partially explained by the fact that one third had hypoxic brain damage caused by etiologies other than cardiac arrest. However, further statistical analysis regarding significance of this finding was not performed due to the small sample size. It remains unclear to what extent RSE, substantial hypoxic brain damage, and discontinuation of life-supporting procedures after intensive care, have contributed to this poor outcome in the sense of a vicious circle, that is, self-fulfilling prophecy (Geocadin et al., 2012). In contrast, reports on brain tumors in RSE patients and associations with outcome are limited. In our cohort, 67% of patients with brain tumors died. This finding contrasts with the data of a study on the detection and treatment of RSE in the ICU, which revealed an improvement of the level of consciousness in all patients with brain tumors on anticonvulsants (Drislane et al., 2008). However, the number and proportion (n = 4; 4%) of patients with brain tumors was small, limiting comparability and generalizability. In a recent study by Swisher et al. (2012) on treatment of RSE in 23 patients with metastatic brain tumors, seizure control was achieved in 70% and mortality was 0%. A clear definition of RSE was not provided, thus limiting comparability. Our results show a different trend, although our small sample size limits generalizability as well. Furthermore, size and localization have substantial influence on epileptogenic activity, outcome (van Breemen et al., 2007; You et al., 2012) and therapeutic decisions. Although all five surviving patients with brain tumors in our cohort had frontal tumor localization, nonsurvivors had additional or other brain regions involved; however, there was no significant difference in tumor size between survivors and nonsurvivors. Compared to our cohort, patients of a recent retrospective study on the acute management of RSE were younger (mean age 57 vs. 62 years in our study) and needed less mechanical ventilation (17% vs. 82%), possibly explained by the fact that the AED regimen was selected intentionally to minimize the need for intubation (Swisher et al., 2012). The remaining, independent association of NCSE in coma with an approximately fourfold risk of death is underscored by studies that previously identified this relation (Rossetti et al., 2006, 2008).
Both generalized convulsive seizures at SE onset, as well as NCSE in come, were associated with an increased risk of death in our cohort. However, after exclusion of patients with HE, the association of generalized convulsive seizures with mortality lost significance, as all patients with refractory SE with generalized convulsions at onset had HE as the underlying seizure etiology.
Duration of RSE was significantly longer in nonsurvivors, an association that has already been described (Towne et al., 1994; Towne, 2007; Drislane et al., 2009). Analysis regarding the relations between RSE duration and the three most common etiologies in our cohort (i.e., HE, brain tumors, as well as known and uncontrolled epilepsy) revealed that differences in seizure duration between survivors and nonsurvivors tended to be more distinct in patients with uncontrolled epilepsy, an etiology that was not shown to have significant high relative risks of death in contrast to brain tumors and HE. Our results suggest that the impact of seizure duration on outcome tends to be less pronounced in the presence of an underlying, severe etiology of RSE. However, this inverse correlation did not reach significance, and sound analyses for the effect of RSE duration in other etiologies could not be performed due to small sample sizes. Similar results have been reported from a retrospective study on the duration of RSE and outcome where seizure duration of <10 h was associated with more favorable outcome, a result that lost significance once etiology, presentation in coma, and SE types were accounted for (Drislane et al., 2009). In addition, case reports and case series report survival with meaningful functional and cognitive recovery, even after prolonged RSE (Cooper et al., 2009; Drislane et al., 2009; Standley et al., 2012). Therefore, seizure duration alone should not be considered a reason to discontinue treatment.
Other clinical characteristics, such as age, the use of intravenous anesthetic drugs, mechanical ventilation, and development of infections during RSE were not significantly related to mortality in our cohort. A recent study on outcome in patients with SE contrasts our findings by describing significant associations of age, mechanical ventilation, and treatment with third-line antiepileptic drugs (including intravenous anesthetic drugs) with poor outcome (Kowalski et al., 2012). This discrepancy may result from the differences in the study populations regarding treatment refractoriness (i.e., RSE in our cohort vs. SE patients) and seizure type (9% CSE in our cohort vs. 92% CSE) as well as possible confounding by indication, as patients with more severe SE may be more likely to receive intravenous anesthetic drugs.