To the best of our knowledge, this is the first report comparing the neurological outcome before and after the implementation of treatment protocols for FCSE in children. This study suggests that treatment protocols for FCSE in children improve the outcomes. In contrast, treatment protocols using fPHT or PB were not associated with better outcomes nor avoidance of anesthetic therapy in children with FCSE.
The efficacy of second-line drugs such as fPHT, valproate, and levetiracetam has been reported to be similar in patients with benzodiazepine–resistant SE . Moreover, we have previously reported that there is no difference in the efficacy of fPHT and MDL ci in pediatric patients with benzodiazepine–resistant SE . Previous studies showed that the efficacy of several second-line agents was similar in patients with benzodiazepine–resistant SE, including febrile SE. However, the effects with or without treatment protocols, or those due to differences in protocols in a real-world setting are yet to be clarified.
4.1 Comparison of the effects of the presence or absence of a protocol (period I vs period II+III)
There are several reports on the usefulness of treatment protocols for SE in children [19, 20]. Xie et al.  reported that before and after the implementation of a protocol, the time between the first detection of seizures and the administration of first- and second-line drugs was significantly reduced. Cassel et al.  also reported that the time to second-line drug administration was significantly reduced by the implementation of a protocol. Herein, the median time from seizure onset to BCT initiation was similar between period I and period II+III. Nevertheless, during period I, there were few patients in whom BCT was started more than 24 h after seizure onset (range, 2.5–85.2 h). In period II + III, BCT was started within 24 h after onset in all patients (range, 2.1–21.3 h). In period I, BCT was started more than 24 h after onset in three patients, and the outcomes of two of these patients were poor. As reported in previous studies, we similarly found that treatment protocols made earlier treatment possible, which may have resulted in better outcomes for the patients in our study.
Although there was no standard protocol in period I, in practice many patients were treated with MDL ci without EEG monitoring to achieve termination of clinical seizures. In a previous study, 164 patients with SE underwent continuous EEG monitoring, and 48% of the patients had persistent electrographic seizures, even after the apparent seizures were suppressed . Nagase et al.  reported that five out of eight children who received MDL ci for refractory FSE had breakthrough seizures, and concluded that clinical seizure control using MDL ci without EEG monitoring is insufficient to prevent neurological damage. Our results might potentially be explained by the presence or absence of EEG monitoring.
The rate of final diagnosis of AE was significantly higher in period I than in period II+III. Tighter control of seizures with EEG monitoring might have suppressed excitotoxicity and the development of AE. Indeed, prolonged seizures cause excitotoxicity , and the association of excitotoxicity with the pathophysiology of AE, such as AESD, has been reported [24, 25].
In the present study, the median patient age was lower in period I than in period II+III. In a systematic review of outcomes from pediatric CSE, younger patients had higher rates of sequelae and mortality . The poorer outcomes of patients in period I than those of patients in period II+III in our study may be due to the younger age of the patients.
4.2 Comparison of the effects of different protocols (period II vs period III)
The outcomes were not different between treatment protocols with and without fPHT or PB. We expected better outcomes in period III than period II, and the use of second-line drugs to reduce the number of cases of BCT, resulting in a reduction of VAP, a complication of mechanical ventilation. However, the use of second-line drugs did not improve outcomes or reduce the VAP incidence.
In period II, the proportion of patients receiving MDL ci was higher than that in period III. However, contrary to treatment in period I, treatment in periods II and III was performed under EEG monitoring. This allowed breakthrough seizures to be detected, which might explain the lack of differences in outcome. Treatment using fPHT occurred more frequently in period III than in period II. As a second-line treatment for benzodiazepine-resistant SE, fPHT is used worldwide [5–8]. However, its effect on febrile seizures is still unknown, and some reports have shown negative effects. Olson et al.  investigated the preventive effects of PB, valproic acid, and phenytoin on hyperthermia-induced seizures in rat pups. Animals injected with PB and valproic acid had seizure temperature thresholds that were significantly higher than those of controls, while animals injected with phenytoin had seizure temperature thresholds that were equal to or lower than those of controls. Among 56 patients experiencing 62 episodes of childhood FSE, phenytoin administration terminated seizures in 9 (14.5%) episodes and failed to terminate them in 25 (40.3%) episodes . Atypical febrile seizures, including SE, are reported to be associated with sodium channel mutations [28–30]. Phenytoin, a sodium channel blocker, may have little effect on FSE . In our study, during period III, fPHT was used mainly as a second-line drug. The lack of difference in outcomes between periods II and III may be due to the poor effect of fPHT on FSE. In recent years, studies have reported that the therapeutic effect of levetiracetam on SE is equivalent to that of phenytoin [31, 32]. Reductions in the severity and mortality of hyperthermic seizures have been observed in levetiracetam-treated rats with cortical dysplasia, where levetiracetam exerted protective effects against hyperthermic seizure-induced blood-brain barrier disruption . In childhood FSE outcomes may be improved by using levetiracetam instead of phenytoin as a second-line drug.
Outcomes of SE are affected by pre-existing neurological abnormalities [34, 35], but a major strength of the present study is that it only targeted patients without pre-existing neurological abnormalities. However, this study also has some limitations. First, it was an observational study without controls, based at a single center. Second, although the data were extracted from a prospectively collected database, the study itself was retrospective. It is unlikely that all of the listed exclusion criteria were documented accurately in every patient. Third, this study targeted febrile seizures, not afebrile seizures such as epilepsy and hypoxic-ischemic encephalopathy. Therefore, the results are not applicable to seizures caused by other factors.