- Top of page
Summary: Background: New continuous infusion antiepileptic drugs (cIV-AEDs) offer alternatives to pentobarbital for the treatment of refractory status epilepticus (RSE). However, no prospective randomized studies have evaluated the treatment of RSE. This systematic review compares the efficacy of midazolam (MDL), propofol (PRO), and pentobarbital (PTB) for terminating seizures and improving outcome in RSE patients.
Methods: We performed a literature search of studies describing the use of MDL, PRO, or PTB for the treatment of RSE published between January 1970 and September 2001, by using MEDLINE, OVID, and manually searched bibliographies. We included peer-reviewed studies of adult patients with SE refractory to at least two standard AEDs. Main outcome measures were the frequency of immediate treatment failure (clinical or electrographic seizures occurring 1 to 6 h after starting cIV-AED therapy) and mortality according to choice of agent and titration goal (cIV-AED titration to “seizure suppression” versus “EEG background suppression”).
Results: Twenty-eight studies describing a total of 193 patients fulfilled our selection criteria: MDL (n = 54), PRO (n = 33), and PTB (n = 106). Forty-eight percent of patients died, and mortality was not significantly associated with the choice of agent or titration goal. PTB was usually titrated to EEG background suppression by using intermittent EEG monitoring, whereas MDL and PRO were more often titrated to seizure suppression with continuous EEG monitoring. Compared with treatment with MDL or PRO, PTB treatment was associated with a lower frequency of short-term treatment failure (8 vs. 23%; p < 0.01), breakthrough seizures (12 vs. 42%; p < 0.001), and changes to a different cIV-AED (3 vs. 21%; p < 0.001), and a higher frequency of hypotension (systolic blood pressure <100 mm Hg; 77 vs. 34%; p < 0.001). Compared with seizure suppression (n = 59), titration of treatment to EEG background suppression (n = 87) was associated with a lower frequency of breakthrough seizures (4 vs. 53%; p < 0.001) and a higher frequency of hypotension (76 vs. 29%; p < 0.001).
Conclusions: Despite the inherent limitations of a systematic review, our results suggest that treatment with PTB, or any cIV-AED infusion to attain EEG background suppression, may be more effective than other strategies for treating RSE. However, these interventions also were associated with an increased frequency of hypotension, and no effect on mortality was seen. A prospective randomized trial comparing different agents and titration goals for RSE with obligatory continuous EEG monitoring is needed.
Refractory status epilepticus (RSE), defined as SE that fails to respond to first- and second-line therapy, occurs in 9–31% of patients with SE (1–3) and is associated with high morbidity and high mortality (4–7). Treatment of RSE has not been studied in a prospective trial, and guidelines give a spectrum of options. Although recent reviews have recommended continuous intravenous (i.v.) midazolam (MDL) (1,8,9) or continuous i.v. propofol (PRO) (10,11) as alternatives to phenobarbital (PB) or continuous i.v. pentobarbital (PTB), evidence for this treatment practice is largely anecdotal.
In accordance with evidence from prospective, double-blind, multicenter studies (2,12), most physicians agree on the use of lorazepam (LZP), with a success rate of 65%, as the initial treatment for SE, followed by phenytoin (PHT) or fosphenytoin as second-line therapy. However, in a recent survey of American neurologists, there was little agreement on third- and fourth-line therapy for RSE (authors' unpublished data). This uncertainty may reflect the fact that no large prospective trial has compared different treatment options for RSE.
In this systematic review, we compared the efficacy and outcome of all published adult RSE patients treated with one of three continuous infusions: PTB, MDL, or PRO. MDL (13,14), PRO (15,16), and PTB (17) all bind to the γ-aminobutyric acid type A receptors (GABAA), augmenting GABAergic transmission, thereby imparting anticonvulsant and sedative–hypnotic properties. All three medications can be given as i.v. infusions. However, their pharmacokinetic properties differ substantially. MDL (a 1,4-benzodiazepine belonging to the 1,2-anelated subgroup) is a fast-acting water-soluble benzodiazepine (BZD) with a half-life of 1.2 to 12.3 h (18,19). PRO (2,6-diisopropylphenol) is a very fast acting agent with a second exponential phase half-life of 34 to 56 min, reflecting its high rate of metabolic clearance, and a third-phase half-life of 184 to 480 min (20). PTB (sodium-5-ethyl-5,1-methylbutyl barbiturate) is a barbiturate with onset of action after 15–20 min and a half-life of 15–60 h (21).
We compared treatment response, complications, and mortality in RSE patients treated with PRO, MDL, or PTB. We also evaluated the efficacy of different treatment intensities as reflected in the EEG titration goal for continuous-infusion antiepileptic drug (cIV-AED) therapy, by comparing EEG background suppression (isoelectric or burst-suppression pattern) (22–24) with that of seizure suppression only (25).
- Top of page
The initial literature search identified 223 studies published between January 1970 and September 2001. We screened 79 articles in detail, after eliminating animal studies and review articles. Of these studies, we excluded 19 articles reporting on the treatment of children only; 12 describing RSE therapy without MDL, PRO, or PTB; six abstracts; four in which patients were initially treated with a cIV-AED not investigated in this study; three that did not meet our criteria for RSE; three in which only single i.v. doses of the investigated AEDs were used; two that did not report immediate treatment response; one study of simple partial SE; and one because of duplicate subject reporting.
One hundred ninety-three RSE patients were identified for further analysis among the 28 studies that met our inclusion criteria (Table 1). Patients were treated with MDL (n = 54) (7,28–36), PRO (n = 33) (5,15,36–44), or PTB (n = 106) (4–6,23–25,45–47)(Table 1). A prespecified treatment protocol was used in 167 cases (43 MDL, 21 PRO, 103 PTB), and in 26 (11 MDL, 12 PRO, three PTB), no protocol was described. Demographic and clinical information is summarized in Table 2.
Table 1. Studies included in the systemic review
|First author||Journal||Year||RSE patients treated with continuous infusion of|
|Young GBa||Can J Neurol Sci||1980||0||0||2|
|Crisp CB||Clin Pharm||1988||1||0||0|
|Campostrini Ra||Nuova Riv Neurol||1991||0||3||0|
|Kumar Aa||Crit Care Med||1992||4||0||0|
|Cortina J||Clin Neuropharmacol||1993||1||0||0|
|McBurney JW||J Epilepsy||1994||0||0||1|
|Borgeat A||Intens Care Med||1994||0||1||0|
|Hantson P||Intens Care Med||1994||0||1||0|
|Merigan KS||Acad Emerg Med||1995||0||1||0|
|Mirski MA||Crit Care Med||1995||1||0||0|
|Total|| || ||54||33||106|
Table 2. Patient characteristics
| ||Continuous i.v. AED||Total (n = 193)|
|Midazolam (n = 54)||Propofol (n = 33)||Pentobarbital (n = 106)|
|Demographics|| || || || |
| Age (yr)||51 ± 21 (54)||47 ± 20 (33)||46 ± 20 (62)||48 ± 20 (149)|
| Female gender||67 (36/54)||64 (21/33)||53 (33/62)||60 (90/149)|
|History of epilepsy||33 (17/52)||28 (8/29)||38 (23/61)||34 (48/142)|
|Primary cause of SE|| || || || |
| Stroke or CNS tumor||28 (15/54)||21 (7/33)||16 (16/102)||20 (38/189)|
| Epilepsy related||19 (10/54)||18 (6/33)||21 (21/102)||20 (37/189)|
| Toxic–metabolic encephalopathy||20 (11/54)||18 (6/33)||19 (19/102)||19 (36/189)|
| CNS infection||17 (9/54)||18 (6/33)||20 (20/102)||19 (35/189)|
| Hypoxia–ischemia||6 (3/54)||12 (4/33)||15 (15/102)||12 (22/189)|
| Traumatic brain injury||6 (3/54)||6 (2/33)||5 (5/102)||5 (10/189)|
| Othera||6 (3/54)||6 (2/33)||6 (6/102)||6 (11/189)|
|Initial seizure type|| || || || |
| GCSE||68 (36/53)||88 (22/25)||89 (70/79)||82 (128/157)|
| NCSE||32 (17/53)||12 (3/25)||11 (9/79)||18 (29/157)|
|Seizure type at the time of cIV-AED|| || || || |
| GCSE||11 (6/53)||24 (8/33)||78 (73/93)||49 (87/179)|
| NCSE||89 (47/53)||76 (25/33)||22 (20/93)||51 (92/179)|
|Total duration of status epilepticus (h)||30.0 (48)||21.5 (28)||13.0 (50)||22.0 (126)|
|APACHE-2 score||20.0 (40)||24.0 (21)||24.0 (6)||20.0 (67)|
|Treatment before ci.v.-AED|| || || || |
| Number of AEDs||3.0 (54)||2.0 (33)||3.0 (62)||3.0 (149)|
| Phenytoin||93 (50/54)||91 (30/33)||95 (59/62)||94 (140/149)|
| Benzodiazepine||76 (41/54)||85 (28/33)||87 (54/62)||83 (123/149)|
| Phenobarbital||57 (31/54)||36 (12/33)||86 (53/62)||64 (96/149)|
| Valproic acid||43 (23/54)||12 (4/33)||5 (3/62)||20 (30/149)|
| Carbamazepine||20 (11/54)||12 (4/33)||8 (5/62)||13 (20/149)|
| Any other AEDb||13 (7/54)||12 (4/33)||15 (9/62)||13 (20/149)|
The majority (82%) of patients had GCSE as the initial seizure type, but 51% had NCSE at some point before starting cIV-AED therapy (Table 2). NCSE was most often treated with MDL, and GCSE, most often with PTB (Table 2). The median duration of SE was longest in MDL-treated (30 h) and shortest in PTB-treated cases (13 h). Primary causes of SE were similar in the three treatment groups.
Treatment characteristics and responses
Most patients were treated with PHT, BZDs, and PB before cIV-AED therapy (Table 2). The mean PHT level before initiation of cIV-AED treatment was 19 ± 9 μg/ml (n = 97); levels did not significantly differ among the three treatment groups. Table 3 summarizes the mean loading doses, minimal and maximal infusion rates, and the duration of cIV-AED therapy. The duration of infusion was longest with MDL (96 h) and shortest with PTB (30 h). In the entire cohort, 15% experienced short-term treatment failure (28 of 193); 25%, breakthrough seizures (37 of 150); 49%, withdrawal seizures (65 of 132); and 10% were changed to a different cIV-AED (17 of 173). Continuous EEG monitoring was performed significantly less frequently in PTB-treated patients (27%) than in patients treated with MDL or PRO (78%) (p < 0.0001; Table 3).
Table 3. Treatment characteristics
| ||Continuous i.v. medication|
| Loading dose (mg/kg)||0.2||1.0||13.0|
| Minimal infusion rate (mg/kg/h)||0.08 ± 0.04||2.94 ± 2.00||1.84 ± 1.59|
| Maximal infusion rate (mg/kg/h)||0.23 ± 0.17||6.98 ± 5.34||3.17 ± 2.11|
|Duration of continuous infusion (h)||96.0 (53)||36.0 (31)||30.0 (61)|
|EEG monitoring|| || || |
| Continuous EEG monitoring||80 (43/54)||76 (25/33)||27 (29/106)|
| Intermittent EEG monitoring||11 (6/54)||15 (5/33)||71 (75/106)|
| None or unknown||9 (5/54)||9 (3/33)||2 (2/106)|
|Titration goal|| || || |
| Seizure control only||100 (43/43)||62 (13/21)||4 (3/82)|
| EEG background suppression||0 (0/43)||38 (8/21)||96 (79/82)|
Effect of cIV agent on treatment response
Among patients treated with MDL, breakthrough seizures (p < 0.001) and change to a different cIV-AED (p < 0.01) were more frequent, and hypotension less frequent (p < 0.001), when compared with the other two medications combined (Fig. 1 and Table 4). PTB treatment was associated with the lowest frequency of short-term treatment failure (p < 0.01), breakthrough seizures (p < 0.001), and change to a different cIV-AED (p < 0.001), whereas hypotension (p < 0.001) was more frequent in these patients (Fig. 1 and Table 4).
Figure 1. Treatment response and outcome in patients with refractory status epilepticus treated with midazolam, propofol, or pentobarbital. Significance was tested with the χ2 or Fisher's exact test of each treatment against the other two. *p < 0.01; **p < 0.001.
Download figure to PowerPoint
Table 4. Treatment responses and outcome
| ||All RSE patients||Nonconvulsive RSE patients|
|Midazolam (n = 54)||Propofol (n = 33)||Pentobarbital (n = 106)||Midazolam (n = 47)||Propofol (n = 25)||Pentobarbital (n = 20)|
|Acute failure||20 (11/54)||27 (9/33)||8 (8/106)a||23 (11/47)||32 (8/25)||20 (4/20)|
|Breakthrough seizures||51 (23/45)b||15 (2/13)||12 (11/92)b||56 (22/39)b||0 (0/6)||0 (0/11)a|
|Withdrawal seizures||63 (25/40)||46 (6/13)||43 (34/79)||66 (23/35)||50 (3/6)||33 (2/6)|
|Hypotension|| || || || || || |
| Requiring pressors||30 (14/47)b||42 (10/24)||77 (79/103)b||30 (13/43)||39 (9/23)||45 (9/20)|
| Refractory hypotensionc||2 (1/47)||8 (2/24)||3 (3/103)||2 (1/43)||9 (2/23)||0 (0/20)|
|ci.v.-AED changed||21 (10/47)a||20 (4/20)||3 (3/106)b||20 (8/41)||25 (3/12)||10 (2/20)|
|Mortality||46 (25/54)||52 (16/31)||48 (49/102)||47 (22/47)||56 (14/25)||30 (6/20)|
In a subgroup analysis of patients with NCSE before cIV-AED administration (Table 4), breakthrough seizures were more frequent with MDL (p < 0.001) and less frequent with PTB treatment (p < 0.01). The number of patients with pure GCSE treated with PRO or MDL was too small to allow meaningful statistical comparisons.
Effect of titration aim on treatment response
Among 167 patients treated with a treatment protocol, 146 specified a titration aim. These were stratified into two groups: (a) “seizure suppression” (43 MDL, 13 PRO, three PTB), and (b) “EEG background suppression” (eight PRO, 79 PTB). Included in this analysis was one study that did not specify a titration goal in the protocol but specifically analyzed the effect of EEG suppression on outcome (48). Patients treated with EEG background suppression were less likely to have breakthrough seizures (4 vs. 53%; p < 0.001), and more likely to have hypotension (76 vs. 29%; p < 0.001; Fig. 2). There was no significant effect of the titration goal on short-term treatment failure, withdrawal seizures, or cIV-AED change.
Figure 2. Treatment response and outcome in patients with refractory status epilepticus stratified according to the titration goal into seizure control only versus EEG background suppression. Significance was tested with the χ2 or Fisher's exact test. *p < 0.05; **p < 0.01; ***p < 0.001.
Download figure to PowerPoint
Mortality and functional outcome
Of the patients who died, 90 (48%) of 187 and only 48 (29%) of 164 returned to their premorbid functional baseline. Patients that died were older (54 ± 20 vs. 42 ± 19 years; p < 0.001), had higher APACHE-2 scores (23 ± 5 vs. 18 ± 6; p < 0.001), had longer median seizure duration (24.0 vs. 12.0 h; p = 0.01), and more often had acute symptomatic SE not related to epilepsy (79 vs. 55%; p = 0.004). Seizure type (GCSE vs. NCSE), gender, hypotension (any, and refractory), and delayed seizure control were not associated with outcome.
- Top of page
Discrepancies between prospective, randomized trails and meta-analyses are well described (49–51), and this problem is particularly common when the data consist of pooled data from case series. However, in the absence of data from a prospective trial for a rare condition, carefully performed systematic reviews do often point in the correct direction (52) and provide concise summaries of the best available evidence (53). The findings of our systematic review cannot substitute for a prospective, randomized clinical trial, but may provide valuable information in the planning phase of such a study. These data are intended not to provide a firm treatment recommendation for RSE patients, but to give a concise summary of experience with these patients.
Outcome was poor overall and was not associated with the choice of cIV-AED (MDL, PRO, or PTB) or the titration goal (titration to seizure suppression or EEG-background suppression). Confirming prior studies of SE, we found mortality to be associated with older age (54,55), etiology (54–59), seizure duration (55,57), and APACHE-2 scores (7).
Because the diseases that cause SE are the most important determinants of mortality and functional outcome, treatment responses may be a more useful end point for comparing AEDs. In this review, treatment responses were most favorable with the use of PTB and when cIV-AED therapy was titrated to EEG background suppression. Specifically, PTB was associated with reduced frequency of short-term treatment failure, breakthrough seizures, and the need to switch to another agent, and titration to EEG background suppression was associated with reduced frequency of breakthrough seizures. NCSE, which is more refractory to therapy than GCSE (3), was substantially more common in MDL- and PRO- than in PTB-treated patients when cIV-AED therapy was started (Table 2), which might explain the superior treatment response seen with PTB. However, we found that PTB was associated with less frequent breakthrough seizures than were the other two agents, even when the analysis was limited to NCSE patients.
Because the vast majority of patients treated with a goal of EEG background suppression were given PTB (79 of 87), it is difficult to determine whether PTB or the titration goal per se might be responsible for the improved treatment response we observed. Another major limitation of the available data is the fact that compared with those treated with MDL or PRO, significantly fewer PTB-treated patients underwent cEEG monitoring. This discrepancy may very likely explain the lower frequency short-term treatment failure and breakthrough seizures found in PTB-treated patients. Without cEEG monitoring, the response of cIV-AED treatment can be difficult to interpret, because subclinical electrographic seizure activity can be detected in 48% of patients after control of convulsive SE (60).
Significantly fewer patients were changed from PTB to another cIV-AED when compared with MDL or PRO. This might reflect the fact that many physicians believe that PTB infusions are the ultimate escalation of RSE therapy, and the fact that at the time many of the PTB cases were treated, MDL and PRO were not yet available as treatment alternatives. Retrospectively, these questions are difficult to address and will have to be answered by a prospective trial.
The significantly higher frequency of hypotension in PTB cases probably reflects the strong cardiovascular depressant effects of this agent. Negative effects of barbiturates on cardiac tone and contractility have been reported as the main disadvantage of barbiturates (4), frequently resulting in cardiac instability (30). Confirming others, we found hypotension to be significantly less frequent with the use of MDL (1) or PRO (61,62). However, PTB was frequently titrated to EEG background suppression, which also was associated with hypotension. Hypotension was usually easily manageable with fluids, pressors, or a temporary decrease in the infusion rates of cIV-AEDs. Refractory hypotension, not easily treatable, was not associated with any one of the cIV-AEDs or titration goals.
Delayed seizure control affects treatment efficacy and mortality (55,57). The pharmacokinetic properties of MDL and PRO (18–21) suggest that these agents may be advantageous for terminating seizures more quickly than PTB. Data in our systematic review were not sufficient to compare this important outcome variable.
In addition to the generally limited conclusions that a systematic review allows, this study is limited by the small numbers of reported cases, the possibility of publication bias, the retrospective nature of its design, the lack of cEEG monitoring in many cases, and differences in ICU management between centers and between reports from the 1980s and 1990s. To minimize these limitations, we rigorously applied stringent inclusion and exclusion criteria.
A prospective, multicenter study should randomize patients to different medications (including PTB in one arm) and treatment protocols (seizure suppression alone vs. more aggressive EEG background suppression). In this trial, patients with NCSE and convulsive SE should be analyzed separately, continuous EEG monitoring must be obtained in all patients, and results should control for patient age, etiology, seizure duration before treatment, and APACHE-2 scores. In combination with neuroprotective strategies, we hope that the results of such a trial will help to improve outcome in this high-risk, critically ill patient population.