Approximately 25% of children presenting with seizures will develop refractory status epilepticus (RSE), defined as ongoing seizures despite the administration of an appropriately dosed benzodiazepine and a second-line anti-seizure medication with a different mechanism of action (1–3). Despite over 4 decades of research, we have not yet determined the optimal management strategy for pediatric RSE.

Continuous seizures have significant effects on the developing brain. Metabolic derangements and widespread neuronal damage begin after as little as 20 min of continuous seizures (4). As seizures continue, gamma-aminobutyric acid (GABA) inhibition declines due to pre-synaptic GABA receptor downregulation (5, 6). This decrease in GABA receptors provides a likely mechanism for why GABAergic medications, including benzodiazepines, become less effective in prolonged seizures, making them difficult to terminate (7–9).

Midazolam is a relatively inexpensive, short-acting benzodiazepine with a half-life of approximately 2 hours when administered as a bolus. Like all benzodiazepines, midazolam exerts its anticonvulsant and sedative effects by binding to the GABA_A receptor and increasing chloride conductance (10). However, unlike other benzodiazepines, midazolam's chemical structure and water solubility permit it to be used in a continuous infusion, and as such, it has become a mainstay in the management of RSE worldwide (10–13).

There is considerable variation in midazolam administration practices regarding starting doses, titration strategies, bolus use, and doses at which treatment is discontinued in favor of alternative third-line agents (14). Such heterogeneity in practice patterns may risk patients suffering from adverse effects of excessive midazolam administration. Alternatively, subtherapeutic doses and delayed initiation of midazolam may be ineffective in other patients and result in a prolonged duration of RSE. A key challenge in pediatric RSE is that midazolam infusions are often initiated late in the clinical course at low doses (1.0 μg/kg/min) with slow titration, likely resulting in lower efficacy from progressive impairment of GABA-mediated inhibition in prolonged seizures (8, 9, 15, 16).

To inform the future development of international protocols and clinical practice guidelines, we sought to address a key gap in the literature on the optimal midazolam prescription strategy. We conducted a systematic review of midazolam infusion strategies in pediatric RSE (ages 1 month to 21 years old), characterizing total seizure duration, midazolam infusion escalation, and the frequency of respiratory depression and hypotension. We hypothesized that the frequency of seizure termination and the incidence of hypotension and respiratory depression would be greater in studies with protocols starting at a midazolam dose of greater than 3.0 μg/kg/min or protocols incorporating more aggressive dose titration, including bolus use with infusion increases (17).

We registered the review protocol on PROSPERO (CRD42023413038) and reported our results according to PRISMA guidelines (Supplementary Data 1) (18).

The objective of this review was to identify the optimal midazolam infusion strategy for pediatric patients with RSE. Eligible studies included randomized control trials (RCTs) and non-randomized studies published from 1990 onward involving infants and children aged 4 weeks or greater (corrected gestational age) up to 21 years, whose RSE was managed with continuous midazolam intravenous infusions (CIVs). Midazolam was infrequently used prior to 1990. Individual case studies, reviews, and editorials were excluded. Abstracts published before 2018 without an accompanying full text were also excluded. Studies including both pediatric and adult participants were included if they contained data for at least 2 pediatric participants with patient-level information.

The intervention group included all patients receiving midazolam infusions starting at 1.0–3.0 μg/kg/min, while the comparison group comprised all patients receiving midazolam infusions starting at 3.1 μg/kg/min or higher. The primary outcome was the frequency of clinical seizure termination, acknowledging that not all centers have access to continuous electroencephalogram (cEEG) monitoring. Secondary outcomes were the time to seizure cessation from midazolam CIV initiation and the frequency of respiratory depression and hypotension. When available, the severity of respiratory depression and hypotension and the interventions required were recorded.

Guided by a professional medical librarian, we searched Medline ALL (Ovid), Embase + Embase Classic (Ovid), Cochrane CENTRAL (Ovid), and Web of Science (Clarivate) using medical subject headings (MeSH), Emtree headings, and keywords (Supplementary Data 2). No studies were excluded based on language. The primary search was completed in March 2023 and updated in February 2024. Studies excluded during full-text review are listed in Supplementary Data 3.

Using Covidence, 3 authors (AA, DA, NM) independently reviewed titles and abstracts in the first stage and full-text articles in the second stage (19). A fourth author (KJ) resolved any disagreements through discussion. Pediatric data were extracted from studies including both adult and pediatric participants, provided that the study had at least 2 pediatric participants with available patient-level data (KJ, NM). All data were independently extracted by the primary authors, with disagreement resolved through discussion. All available results for each outcome measure were recorded.

We evaluated the risk of bias using a modified Cochrane Risk of Bias 2 tool (RoB2) for RCTs and the Risk Of Bias In Non-randomised Studies- of Interventions (ROBINS-I) tool for non-randomized studies (20, 21). All risk of bias assessments were performed for the primary outcome (frequency of seizure termination) independently by the primary authors (KJ, NM), with disagreements resolved through discussion. Narrative data synthesis was performed due to heterogeneity in RSE definitions, patient populations, and treatments. Descriptive statistics were performed using R Studio (R version 4.3.1) (22, 23).

We conducted a comprehensive systematic review of the literature to determine the optimal prescription strategy for pediatric RSE, aiming to better define an efficacious starting dose and titration strategy. We found 19 studies worldwide, including 3 RCTs and 16 observational trials involving 448 children with RSE treated with midazolam CIVs. Despite significant heterogeneity in starting doses and midazolam titration techniques, the mean effective dose for seizure cessation clustered between 2.0 and 5.0 μg/kg/min in 15/19 studies (Figure 2), with 80% efficacy (321/399). Midazolam-related adverse events included the requirement for mechanical ventilation in 42/221 (19%) children and hypotension requiring IV fluids or no intervention in 18/221 (8%).

Our review critically evaluates the evidence supporting midazolam infusions in pediatric RSE, with greater emphasis on dosing and bolus strategies than previous reviews. We included all studies published from 1990 onward, capturing diverse practices worldwide. While comprehensive, our review has limitations that may decrease its direct applicability to clinical practice and guideline development. Primarily, the clinical and molecular understanding of SE has changed considerably since the late 1980s to early 1990s when midazolam first entered practice (1, 2, 14, 44). Definitions of SE have evolved from including minimum seizure durations to emphasizing early aggressive seizure management (45). This means that patients with RSE in historical studies may have longer seizure durations. Importantly, variations in how RSE is defined today may limit comparison between contemporary studies (1). Across the literature, studies also use differing endpoints and access to cEEG monitoring may impact the rapidity of infusion titration and treatment durations.

Consensus guidelines for the treatment of RSE and pediatric RSE are limited by a lack of evidence. In 2020, the American Epilepsy Society Treatments Committee published a comprehensive review evaluating the efficacy of 8 anticonvulsant medications in treating refractory status epilepticus. They concluded that there is limited evidence to suggest that any one medication is more efficacious than others in seizure termination (14). Our systematic review attempts to strengthen the limited evidence for midazolam by evaluating the impact of midazolam bolus administration and the titration rate on the efficacy of seizure cessation. Unfortunately, across the 19 studies, a great deal of heterogeneity existed regarding infusion strategies and whether boluses were given with each dose titration. These variations and lack of key details limit dosing strategy comparison and identification of any subgroups that could benefit from alternative approaches. Bolus use may greatly impact the time to seizure cessation (46). Luchette et al. performed computer-based simulations of midazolam pharmacokinetics and found that infusions without boluses took longer to reach therapeutic dosing, delaying seizure cessation by 30 min (47). In addition, the use of different classification schemes for seizure etiology across studies precluded secondary subgroup analysis to determine whether certain seizure etiologies are more likely to respond to midazolam.

Beyond inter-study heterogeneity, the development of evidence-based guidelines is further limited by low certainty of evidence. All 19 studies included in this review were at probably high to critical risk of bias. Risk of bias concerns arise from subjective outcomes such as clinical determination of seizure cessation, lack of intervention blinding, and confounding effects due to the co-administration of other anti-seizure medications either for sedation or seizure management. Evaluation of clinically important respiratory adverse events is challenged by confounding factors, including concurrent or previous receipt of anti-seizure medications and studies that intubate per protocol. The incidence of midazolam-related hypotension similarly remains unclear due to confounding from phenobarbital or other anti-seizure medication administration. Limited information is provided regarding the timing of hypotension in relation to midazolam bolus and infusion escalation. Reassuringly, Morrison et al., who used a rapid midazolam escalation strategy, reported only 4 patients with hypotension (33). Two of these patients had transient hypotension associated with phenytoin and midazolam co-administration, which was corrected by 10 ml/kg of IV fluids, while 2 were already receiving vasoactive medication prior to midazolam titration without any increase in their dosing during midazolam titration. Finally, we note the potential bias and limitations of including case series as observational trial data.

With the potential decrease in efficacy of midazolam in prolonged seizures from GABAergic downregulation, it is tempting to abandon midazolam in favor of alternative agents. However, the lack of standardized protocols, questionable use of boluses, and variable data reporting limit conclusions. Despite the above limitations, our review suggests a therapeutic window for midazolam CIVs between 2.0 and 5.0 μg/kg/min and indicates that starting the infusion within this therapeutic window may achieve earlier seizure cessation. Given that midazolam is a relatively inexpensive and widely available medication with potentially fewer respiratory and hemodynamic consequences than phenobarbital or propofol, we advocate for further research to address important dosing and efficacy questions (38, 48). Midazolam may be more effective when used earlier and rapidly titrated with boluses. Our systematic review highlights the vast differences in approaches to continuous midazolam infusions in treating RSE but also differences in how SE and RSE are defined and how seizure etiology is classified. Without consensus on these definitions and standardization of boluses within protocols, comparison across studies and collaboration across countries will remain challenging. Research collaboratives, such as the Pediatric Status Epilepticus Research Group, will play a key role in achieving this standardization (1). Research investigating midazolam and other third-line anesthetics will also need to consider the timing of initiating these medications and clinician hesitation to use anesthetics, including at higher doses, for fear of respiratory depression, hypotension, or potential increased risk of mortality (49).

Midazolam is a commonly used anesthetic infusion for the treatment of pediatric RSE, yet there is limited evidence guiding treatment protocols and clinical practice. Data from 19 studies involving 448 children illustrate significant heterogeneity in infusion dosing, bolus use, seizure termination effectiveness, and incidence of respiratory and hemodynamic adverse events. Our findings suggest a potential therapeutic window at rates of 2.0–5.0 μg/kg/min (0.12–0.30 mg/kg/h), with limited adverse risks. Earlier seizure cessation may occur by targeting this therapeutic window by initiating treatment at higher doses than the typically used 1.0 μg/kg/min (0.06 mg/kg/h) or by rapidly escalating the dose.