Parts of this chapter have been reused from Lyttle et al.¹ This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY-NC-ND 4.0) license, which permits others to copy and redistribute the material in any medium or format, provided the original work is properly cited. See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

To the best of our knowledge, the EcLiPSE trial is the largest and most clinically pragmatic RCT to compare levetiracetam with phenytoin in the treatment of paediatric CSE unresponsive to first-line treatment. This trial, powered for superiority, did not detect a statistically significant difference in any of the primary and secondary outcomes. The direction of effect favoured levetiracetam across the primary outcome (i.e. time from randomisation to CSE cessation) and secondary outcomes (e.g. the need for RSI and SARs), other than for the secondary outcome of admission to critical care, for which the direction of effect favoured phenytoin. These findings were robust in all sensitivity analyses.

The results of the EcLiPSE trial were published in May 2019.¹ A similar RCT, the Convulsive Status Epilepticus Paediatric Trial (ConSEPT), which was undertaken in 13 EDs in New Zealand and Australia, of 233 children aged 2 months to 16 years and used a similar protocol, including identical doses and rates of administration of the two drugs, was published simultaneously in the same journal.⁴¹ The primary outcome of the ConSEPT trial was clinical cessation of seizure activity 5 minutes after the completion of infusion of the study drug, with levetiracetam infused over 5 minutes and phenytoin infused over 20 minutes. Analysis was by intention to treat. Results showed that clinical cessation of seizure activity 5 minutes after completion of infusion of the randomised drug occurred in 68 (60%) patients in the phenytoin group and 60 (50%) patients in the levetiracetam group (risk difference –9.2%, 95% CI –21.9% to 3.5%; p = 0.16). The authors⁴¹ concluded that levetiracetam was not shown to be superior to phenytoin for the second-line management of paediatric CSE.

More recently, the results of the Established Status Epilepticus Treatment Trial (ESETT) from the USA showed no statistical significance difference in its primary outcome (i.e. cessation of status and improvement in consciousness at 60 minutes without the use of additional antiseizure medication) between levetiracetam (at a dose of 60 mg/kg), sodium valproate and fosphenytoin (a pro-drug of phenytoin and at a dose of 20 mg/kg). The primary outcome was achieved in approximately 50% of patients and the frequency of AEs was similar in the three drugs.^70,71

Prior to the publication of the ConSEPT trial, reported CSE cessation rates for levetiracetam and phenytoin were broadly similar to a number of previously reported observational retrospective, and predominantly adult, studies.^14,30,31 However, CSE cessation rates as high as 85–95% have been reported, although these studies display significant heterogeneity in design and outcomes.^72–74 A recent prospective RCT⁷⁴ of only 50 children recruited over a 6-month period reported that levetiracetam (at a dose of 30 mg/kg) terminated CSE in 92% of children and fosphenytoin (at a dose of 20 mg/kg) terminated CSE in 84% of children (p = 0.66).⁷⁴ These rates are considerably higher than those found in the EcLiPSE trial, ConSEPT trial⁴¹ and ESETT.^70,71 This also applied to the time to terminate CSE from the time of commencement of the infusion. Fosphenytoin terminated CSE earlier (2.5 ± 1.4 minutes) than levetiracetam (3.3 ± 1.2 minutes; p = 0.03).⁷⁴ The equivalent median times in the EcLiPSE trial, from time of commencement of infusion, were 17.5 and 24.5 minutes for levetiracetam and phenytoin, respectively, and in the ConSEPT trial these were 17 and 22 minutes, respectively. The methodology of the study by Senthilkumar et al.⁷⁴ was limited and each treatment group comprised only 25 patients (this may explain their markedly discrepant findings with the EcLiPSE trial for both CSE cessation and the speed with which this was achieved). Finally, although fosphenytoin can be administered more rapidly than phenytoin, it can still not be administered as quickly as levetiracetam. Consequently, the findings of Senthilkumar et al.⁷⁴ remain difficult to understand and it is uncertain if they would be generalisable.

One RCT,²⁸ undertaken in adults with CSE, compared the efficacy of i.v. phenytoin (20 mg/kg), sodium valproate (30 mg/kg) and levetiracetam (25 mg/kg) in 150 patients unresponsive to i.v. lorazepam. CSE stopped in 34 (68%) patients treated with phenytoin, 34 (68%) patients treated with valproate and 39 (78%) patients treated with levetiracetam (p = 0.44). A paediatric RCT,⁷² published in 2018, evaluated 100 children aged 3–12 years receiving levetiracetam (30 mg/kg) or phenytoin (20 mg/kg) if CSE continued after one dose of i.v. diazepam.⁷² Efficacy was high and almost identical in both groups. A lower diastolic blood pressure was recorded in phenytoin-treated patients (p = 0.023). It is difficult to translate these findings to clinical practice because of the trial’s design, including its many exclusion criteria and its primary outcome,⁷¹ which was ‘absence of seizure activity within 24 hours’. This is an unusual and very rarely used primary outcome in other studies of CSE. It is also not a practical and ‘real-life’ clinical goal, as the emphasis should be on the termination of the presenting seizure as soon as possible after treatment has been given, and not the child’s condition after 24 hours. Consequently, it would be difficult, and probably inappropriate, to use the same primary outcome in routine clinical practice in the UK and elsewhere. Childhood CSE management in the UK follows the APLS algorithm,¹¹ which is applicable to the vast majority of children presenting to an ED. Our study design used eligibility criteria that were as inclusive as possible, and followed a well-recognised treatment pathway that reflected routine clinical practice.

Children with focal and generalised CSE were included because their management is the same in the APLS algorithm. In addition, it may be difficult to accurately distinguish focal and generalised convulsive seizures in infants and children aged < 3 years.

We did not detect a statistically significant difference between levetiracetam and phenytoin in time to CSE cessation. A superiority design was selected for three reasons: (1) the reported CSE cessation rates for each drug, hypothesising that levetiracetam would be more effective; (2) the absence of any RCT data comparing the effectiveness of either phenytoin or levetiracetam with placebo and (3) the shorter infusion time of levetiracetam (i.e. 5 minutes vs. at least 20 minutes for phenytoin). We selected time from randomisation, and instructed sites to undertake randomisation at the latest possible point that would allow reconstitution of the allocated treatment to provide scientific and clinical rigour. As the median time to commencement of infusion exceeded 10 minutes in each arm, we also undertook a sensitivity analysis, using time to cessation of CSE from commencement of the infusion. This supported our primary analysis findings and demonstrated that the median time from commencement of the infusion was similar to the median time from randomisation in both treatment groups. This is interesting, as it might have been expected that CSE would have been terminated more quickly with levetiracetam (with infusion time of 5 minutes) than phenytoin (with infusion time of at least 20 minutes). One explanation for this observation could be that the anticonvulsant effect of phenytoin may be achieved prior to completion of its infusion. As far as we are aware, there is no literature that has specifically evaluated how rapidly an infusion of phenytoin might terminate CSE once it has been commenced. The ConSEPT trial⁴¹ also showed no significant difference between levetiracetam and phenytoin in time to CSE-cessation.

Progression to RSI in CSE may be required for one or a combination of reasons, including continuing CSE, respiratory depression, clinical deterioration and stability for transfer, or to safely undertake investigations, specifically neuroimaging. However, RSI abolishes visible CSE activity and may, therefore, prevent an assessment of CSE cessation directly related to trial treatment. Participants were, therefore, censored at the time of RSI, but the censoring time was increased to allow for this to be a negative and potentially informative outcome. This may have artificially inflated the time to CSE cessation. However, sensitivity analyses that censored patients at the time of RSI, and defined RSI as a competing risk, did not change our findings.

Safety profiles were similar across both treatments. Owing to their relative infrequency in relation to the trial population size, together with good clinical management in participating sites, the trial showed low rates of SARs. Only 4 out of 286 (1.4%) patients experienced a total of five SARs (three in the phenytoin-treated group and one in the levetiracetam-treated group) and in only one of these patients was the reaction (i.e. marked hypotension) considered as being ‘probably’ related to the study medication (phenytoin). One SAR occurred in a patient who received both anticonvulsants. One phenytoin-treated patient experienced a SUSAR, which manifested as a large increase in seizure frequency and marked sedation within 24 hours of receiving phenytoin. An equally good safety profile was also reported by the ConSEPT trial team,⁴¹ with one death in the phenytoin group 27 days after randomisation because of haemorrhagic encephalitis that was considered to be unrelated to the study drug. The authors reported no other SAEs or SARs. The safety profile of the 255 children (aged 2–17 years) who participated in the ESETT^70,71 was also good. Two deaths occurred, one in the levetiracetam-treated group and one in the sodium valproate-treated group. Life-threatening hypotension and cardiac arrhythmias were rare and did not differ by treatment group in any age. The only significant safety outcome was seen in children requiring intubation more frequently in the fosphenytoin-treated group, an observation that could not be readily explained by the authors. The authors reported no other differences in safety outcomes.⁷¹

The good safety profile of both anticonvulsants is encouraging, particularly for phenytoin, in view of its well-recognised serious potential adverse effects of hypotension, cardiac arrhythmias and severe extravasation reactions, including the ‘purple glove syndrome’.^15,16 Rarely, the arrhythmia may be fatal caused by non-resuscitatable cardiac asystole.¹⁸ In the USA, fosphenytoin, a pro-drug of phenytoin, replaced phenytoin as the preferred second-line management of CSE, primarily because of its slightly faster rate of infusion but also because of its perceived better safety profile. However, it may also cause SARs and this, together with its relative cost to phenytoin, has precluded its use in the UK.

The literature on the safety of levetiracetam is less extensive because of the comparative short period that it has been used in the treatment of CSE. However, despite the fact that this period spans < 20 years, in comparison with the 50 years with phenytoin, it is important to note that, to date, there have been no reports of severe Stevens–Johnson syndrome, purple glove syndrome or fatal cardiac arrhythmias associated with the use of i.v. levetiracetam.

In the EcLiPSE trial, levetiracetam was well tolerated at an infusion rate of 5 minutes and this was more rapid than previously reported (i.e. 10–15 minutes).^27,30,31 Agitation was the most commonly reported AE in the levetiracetam-treated group, as reported previously.²⁶ There were no new or unexpected SARs with levetiracetam. Sedation, somnolence and dizziness are rare side effects in adults, but these may in part reflect the prior use of benzodiazepines or craniotomy in these study populations.^20,74 Anxiety has also been reported in adults, but was not reported as an AE or AR in this or other paediatric studies. It is possible that anxiety in the adult may equate to agitation in the child. Clearly, in view of the fact that 90% of our study population was aged ≤ 10 years (and 41% aged ≤ 2 years), anxiety might be difficult, if not impossible, for them or their carers to describe, and instead used the terms ‘agitated’ or ‘irritable’.

More participants in the levetiracetam-treated group required admission to critical care (either a high-dependency unit or a PICU), but this did not reach statistical significance. This finding is difficult to explain. It was not explained by the demography of the two treatment groups, ongoing CSE, the need for RSI, additional anticonvulsants, or the frequency of AEs or AEs. Potential explanations could include a different type or nature (e.g. more severe) of epilepsy in the levetiracetam-treated group or that the attending clinicians had a lower threshold of transferring participants to critical care, as levetiracetam was a relatively new anticonvulsant. There did not appear to be a difference in the two treatment groups regarding the type of epilepsy or additional comorbid conditions, or the numbers that were still on critical care 24 hours after randomisation following post hoc analysis.

Children who were already receiving levetiracetam and phenytoin as oral maintenance antiepileptic drugs were included in the study because this reflects real life. In most emergency situations and, in particular, in the absence of information from carers, children will be treated in accordance with the APLS algorithm. In addition, a recognised cause of CSE in patients with epilepsy at all ages is poor adherence to antiepileptic medication and consequent low blood levels of the medication. The relatively high proportion of children receiving levetiracetam was partly expected and reflects current clinical practice and perception that this drug has a broad spectrum of action (in treating different seizure types) and is safe. The very small number of children receiving phenytoin was also predictable, as it is perceived as having a very narrow spectrum of action, numerous drug interactions, significant long-term side effects and is difficult to monitor. We consider it unlikely that the inclusion of children on levetiracetam and phenytoin significantly affected our findings. Theoretically, it might have been predicted that those children already receiving levetiracetam would not have responded as well to i.v. levetiracetam or have shown more adverse side effects, or both, although this was not reflected in the overall results. However, the small number of patients receiving levetiracetam across both treatment arms precluded any formal subgroup analysis.

The EcLiPSE trial is a unique trial for many reasons. First, to the best of our knowledge, it is the first adequately powered RCT to compare the efficacy and safety of two anticonvulsants as second-line treatments for CSE. Second, to the best of our knowledge, it is the first adequately powered RCT to evaluate phenytoin as a second-line treatment for CSE, despite this drug’s position as the first-choice second-line treatment for > 50 years. Third, the trial incorporated a nested consent study that evaluated the process of RWPC in a PEM trial.^43,44 Last, to the best of our knowledge, it was the first multicentre RCT to be supported by and delivered across the then emerging PERUKI collaborative.^37,38

This trial has a number of strengths. First, it evaluated a specific step (i.e. second-line treatment) in a UK clinical algorithm for the management of childhood CSE.¹¹ A similar trial⁷⁵ that assessed the first-line non-i.v. treatment of CSE in the same algorithm led to a change in national clinical practice. Second, it demonstrated that RWPC is acceptable and successful, with 385 out of 404 (95%) randomised participants providing consent. Likewise, in those who were randomised and treated, 286 out of 311 (92%) participants provided consent. RWPC is essential for the successful delivery of paediatric emergency care trials. The high consent rate mirrors that found in a previous trial⁷⁵ of first-line CSE management (consent rate 97%) and a pilot RCT⁵⁶ that compared fluid boluses in shock (consent rate 100%). Third, it was a pragmatic trial and recorded only key primary and secondary outcomes in the resuscitation room. This approach, supported by focused data-collection materials and simple allocation and enrolment methods, facilitated successful delivery of the study across all sites, as shown by the small numbers of missed patients, high protocol adherence and accurate data capture for key outcomes. Finally, the trial was conducted in EDs in secondary and tertiary institutions throughout PERUKI, thereby facilitating dissemination and increasing generalisability of our findings.

This trial has some limitations. First, it was open label. A double-blind design was considered too complex for most participating sites, in part because of the markedly different infusion rates of the two drugs, and within the context of the life-threatening and time-critical nature of CSE. Second, the number of participants in one arm of the trial fell below the sample size calculation requirement (134 vs. 140), whereas in the other arm it surpassed requirements (152 vs. 140); however, the effect of this on power is considered unimportant. Third, there was probably subjectivity in the assessment of ‘cessation of all signs of continuous, rhythmic clonic activity’ as the clinical event for our primary outcome, rather than fixed time points to assess CSE cessation. Clearly, these three limitations may collectively increase the risk of bias. However, continual assessment of a child’s condition reflects ‘real-life’ practice in a dynamic situation, in which clinicians constantly evaluate and prepare for the next step in the treatment algorithm. Site training included a simulated demonstration of the end point to ensure an understanding of the key outcome measure for the trial. In the ConSEPT trial,⁴¹ the primary outcome assessment (i.e. CSE cessation 5 minutes after completion of the randomised infusion) was video-recorded. This was to explore possible observer bias owing to the unblinded nature of the study design. Recordings were obtained for 84 (71%) participants in the levetiracetam-treated group and 71 (62%) participants in the phenytoin-treated group. Although this might have been feasible in a research setting in a few sites, it would not be easily applied to routine clinical practice. It would not have been feasible or pragmatic for each participant to undergo a video-recording or an electroencephalogram (EEG) to determine CSE cessation time more precisely. It is not possible to state definitively and without EEG whether or not any patients may have developed non-CSE following either treatment. However, treatment algorithms for non-CSE generally follow the same flow as CSE, and there was no difference between treatment groups in the number of additional anticonvulsants given after trial treatment. EEGs were not used to determine or confirm seizure cessation in the ConSEPT trial⁴¹ or ESTT.^70,71 Third, the time point of randomisation resulted in CSE terminating prior to administration of trial treatment in a number of cases; however, this affected both treatment arms equally, and was essential to maintain high standards of clinical care and avoid treatment delays. Fourth, we included safety measures as key secondary outcomes because of previous reports of harm. However, this trial was not powered to demonstrate difference in SARs (a secondary outcome) between treatment groups, given their low incidence rate. Finally, we considered a superiority design was more appropriate for reasons given above.