The search strategy yielded 433 results (425 MEDLINE, seven congress abstracts, one unpublished study). After excluding duplicate publications and case reports, 55 articles were provisionally selected. We excluded 25 studies after reading the full published papers; thus, 30 studies contributed to this review: the earliest was published in 1993 and the most recent in 2012.
Overall, we identified through a comprehensive search in MEDLINE 25 publications in which the efficacy of intravenous valproate in SE had been reported [29–53]. Cross-referencing identified seven additional reports in congress abstracts [54–60]. Three of these were redundant [55, 57, 60] as they corresponded to case series subsequently described in full publications as part of larger cohorts [35, 40, 45]. Finally, the manufacturer of valproate provided the clinical study report of an unpublished study .
Of the 30 unique studies identified, six were randomized controlled trials, none of them blinded, comparing intravenous valproate to phenytoin, diazepam, or phenobarbital [41–43, 46, 52, 53], and four non-randomized controlled studies comparing intravenous valproate to phenytoin or levetiracetam [48–51]. One randomized controlled trial also included patients with acute repetitive seizures not entailing the diagnosis of SE. For this study, information from patients with SE was obtained from a systematic review . Of the uncontrolled studies, eight were prospective observational studies and 12 were retrospective case series.
Randomized Controlled Studies
Three randomized controlled studies compared intravenous valproate with intravenous phenytoin [41, 42, 46], two with intravenous diazepam [43, 52], and one with intravenous phenobarbital . These studies included 361 patients (183 were randomized to receive valproate; Table 1; Fig. 1). The phenobarbital study  and one diazepam study  included only children; one phenytoin study  included only adults, whereas the remaining studies [41, 42, 52] included all age groups. The most frequent SE was generalized convulsive. Valproate was administered most often at an initial bolus dose of 20 or 30 mg/kg (Table 2).
Non-Randomized Controlled Trials
Non-randomized controlled studies were included in the present review. Two studies utilized a prospective design [49, 50], and two were retrospective [48, 51]. These studies included 455 subjects (197 were randomized to valproate; Table 1).
Prospective Open-Label Studies
The eight prospective studies identified are presented in Table 1. Overall, these included 191 patients with SE treated with intravenous valproate (135 adults, 46 children, and ten of unspecified age). These studies all evaluated intravenous valproate administered in monotherapy, usually administered as a bolus (15–40 mg/kg) followed by a maintenance infusion.
Retrospective Case Series
We identified 12 retrospective case series (Table 3). These included 76 adults, 45 children, and 135 subjects of unspecified age with SE. Again, valproate was usually administered as a bolus (15–40 mg/kg) followed by a maintenance infusion.
Response rates are presented in Table 4. These rates varied across studies, between 35 and 100 % (median 74.35 %). The overall response rate in all studies was 70.9 % (601/848; 95 % confidence interval [CI] 67.8–73.9). Outcome in children (81.3 %; 104/128; 95 % CI 74.5–88) was better than in adults (68.5; 207/302; 95 % CI 63.3–73.8). Response rate was lower in controlled non-randomized studies (61.8 %, 114/185; 95 % CI 54.6–66.6) than in randomized controlled studies (75.4 %; 138/183; 95 % CI 69.2–81.6) or retrospective case series (68.2 %; 197/289; 95 % CI 62.8–73.8). There was no difference in response rates in convulsive (69.7 %; 246/353; 95 % CI 64.9–74.5) compared with non-convulsive SE (75 %; 15/20; 95 % CI 56–94), although the comparison is limited by the small number of studies in which outcome was presented according to SE type.
Comparative Randomized Studies
Six studies have compared the efficacy of intravenous valproate with that of intravenous phenytoin, diazepam, or phenobarbital. Two studies compared valproate with phenytoin in early SE [43, 46], whereas one study compared valproate with phenytoin in established SE .
In the first study  comparing the rate of response, defined as seizure cessation at the end of the infusion, it was significantly (p = 0.046) higher in the valproate group (66 %; n = 35) than in the phenytoin group (42 %; n = 33). Non-responders were switched to the other treatment, where 15 of 19 non-responders to phenytoin (79 %) responded to valproate compared with 3 of 12 non-responders to valproate (12 %) who responded to phenytoin (p = 0.004). Seizure freedom at 24 h was obtained in 29 patients, irrespective of treatment group or sequence. However, the use of a maintenance AED regimen after the bolus infusion was not reported in this study.
In the second study , treatment success was defined as clinical seizure cessation after infusion without need for rescue medication within 20 min of the infusion. No statistically significant difference in the rate of treatment success was found among patients treated with valproate (13 of 18; 72.2 %) and phenytoin (seven of nine patients; 77.8 %). Seizure freedom at 24 h was obtained in all patients (100 %) in both groups.
In the study comparing valproate with phenytoin in established SE, treatment success was defined as cessation of all motor and electroencephalography (EEG) seizure activity within 20 min of starting drug infusion and no return of seizure activity in the following 12 h . According to this criterion, 44 patients receiving valproate (88 %) and 42 receiving phenytoin (84 %) experienced treatment success. Success rates were higher in patients in whom SE had been treated within 2 h (p < 0.05), with only one such patient treated with phenytoin failing to respond.
In the study comparing valproate with diazepam, both in children and in adults, generalized convulsive SE was controlled in 56 % (20 of 36 patients) of the diazepam group and 50 % (15 of 30 patients) in the valproate group . Relapse of SE within 24 h occurred in 25 % (5 of 20 patients) of the diazepam and 20 % (3 of 15 patients) of the valproate group. Both failed to reach statistical significance, suggesting no differences between groups in terms of efficacy.
A recent systematic review assessing the role of valproate in generalized convulsive SE included the above-mentioned studies comparing intravenous valproate with intravenous phenytoin [41, 42, 46] in a meta-analysis . Compared with phenytoin, intravenous valproate had no statistically significant differences in seizure cessation after drug administration (risk ratio [RR] 1.31, 95 % CI 0.93–1.84) and in seizure freedom at 24 h (RR 0.96, 95 % CI 0.88–1.06).
In the study comparing valproate with diazepam in children, SE was successfully controlled in 16/20 patients receiving valproate (80 %) and 17/20 patients receiving diazepam (85 %) . The median time to abrogation of seizure activity was shorter (p < 0.001) in the valproate (5 min) than in the diazepam group (17 min). Breakthrough seizures during the maintenance phase occurred in eight patients in each group.
In the study comparing intravenous valproate with intravenous phenobarbital in children, treatment success was defined as cessation of all convulsive activity within 20 min of anticonvulsant infusion . Seizure termination was significantly higher among patients treated with intravenous loading of valproate (27 of 30 patients; 90 %) than in patients treated with phenobarbital (23 of 30 patients; 77 %) (p = 0.189). In the phenobarbital group, a statistically significant higher relapse of SE within 24 h was observed (12 of 23 children) compared with valproate treatment (4 of 27 children).
Comparative Non-Randomized Studies
Four studies have compared the efficacy of intravenous valproate with that of intravenous phenytoin or levetiracetam [48–51].
One prospective study compared intravenous valproate with intravenous levetiracetam in adults with refractory SE at a tertiary center in India . Refractory SE was defined as seizures lasting for more than 1 h, and patients had to have already received intravenous lorazepam and phenytoin. Clinical seizure cessation after infusion was obtained by 26 of 41 patients in the valproate group (68.3 %) and 28 of 41 patients in the levetiracetam group (73.2 %), although the difference was not statistically significant. Seizure control was not reached by 13 (31.7 %) patients in the valproate group and 11 (26.8 %) in the levetiracetam group (difference not statistically significant), and patients required intubation and administration of propofol or midazolam infusion.
One retrospective study was conducted in adults with convulsive SE who were treated with intravenous valproate as first- or second-line AED ; 12 (24 %) and 37 (76 %) patients were treated with intravenous valproate and phenytoin as first-line treatment, respectively. Treatment success was defined as the cessation of seizures by either clinical or EEG evidence. SE ceased in 9 of 12 patients (75 %) with intravenous valproate and 17 of 37 patients (46 %) with intravenous phenytoin as first-line treatment (difference not statistically significant), whereas it ceased in 7 of 20 (35 %) patients who received second-line valproate.
Kalita et al.  conducted a prospective study in children and adults with convulsive and non-convulsive SE, comparing first-line intravenous valproate with first-line intravenous phenytoin. A total of 65 patients received valproate and 52 received phenytoin as a first-line AED, which resulted in control of SE in 56 % of patients with valproate and 44 % of patients with phenytoin. Crossing over of the treatment in uncontrolled patients resulted in seizure cessation in 41 additional patients, whereas 35 (30 %) remained refractory to both valproate and phenytoin. Within 24 h, seizures recurred in six patients whose SE initially responded to the first or second AED (no further information on valproate or phenytoin was available in the published article).
A retrospective study compared intravenous valproate with phenytoin or levetiracetam as second-line treatment in adults . The primary outcome was the failure of the second-line AED, defined as the need to introduce a further compound to control SE; 15 of 59 patients in the valproate group (25.4 %) failed to control seizures, compared with 29 of 70 patients in the phenytoin group (41.4 %) and 28 of 58 patients in the levetiracetam group (48.3 %).
Status Epilepticus (SE) Subtypes
Certain studies have reported seizure outcome in response to intravenous valproate treatment in defined seizure types.
Generalized Tonic–Clonic SE
The largest group of patients for whom seizure outcome has been reported by seizure type corresponds to patients presenting with generalized tonic-clonic SE. Specific outcome was reported in ten studies [29, 30, 33, 34, 36, 40, 45, 47, 49, 52], including a total of 166 patients. Most subjects (119/166) responded to treatment, yielding an overall response rate of 71.7 % (95 % CI 68.3–75.1).
The other type of SE in which intravenous valproate has been most frequently evaluated is simple or complex focal SE. Overall, data have been reported for 107 patients, included in nine studies [29–31, 33, 34, 36, 45, 47, 57]. Most subjects (83/107) responded to treatment, with an overall response rate of 77.6 % (95 % CI 75.3–79.9).
Information on the effectiveness of monotherapy with intravenous valproate in absence status is limited. Such patients have been reported in six studies [29, 32, 33, 36, 44, 57]. The number of subjects in these studies was low (16 overall, ranging from one to five patients per study). In all studies except one , including three patients in whom absence status was described as atypical, SE was rapidly abrogated by valproate treatment in the majority of cases (12/16 cases). The overall response rate in absence status was 75.0 % (95 % CI 53.8–96.2). In one study, two patients with absence status received valproate as first-line therapy rather than after failure of intravenous benzodiazepines . Both of these patients responded to treatment.
Other Presentations of SE
Three patients with myoclonic SE were studied by Giroud et al.  and four such patients by Jha et al. . One patient in each study failed to respond. The overall success rate was thus 71 %. Patel and Jha  reported ten patients with post-anoxic myoclonic SE occurring after surgery. SE was abrogated successfully in six of these patients (60 %) with intravenous valproate given at a dose of 30 (four patients) or 40 (two patients) mg/kg within 2–10 h. In addition to these patients, two cases of successful treatment of myoclonic SE have been reported by Sheth and Gidal . Überall et al.  reported two cases of status associated with infantile spasms; both responded successfully to intravenous valproate therapy.
Seizure Outcome and Order of Valproate Administration
Two randomized [41, 46] and one non-randomized  controlled trials used valproate as first-drug antiepileptic treatment instead of benzodiazepines. Four randomized [42, 43, 52, 53] and three non-randomized [48, 49, 51] controlled trials used valproate after failure of first-line benzodiazepines (lorazepam or diazepam).
Among studies using valproate as second-line drug, SE was controlled in 50–90 % of patients allocated to valproate in randomized controlled studies, and in 56 % of patients treated with valproate in the non-randomized controlled trial. Among studies using valproate as first-line drug, SE was controlled in 66–72.2 % of patients allocated to valproate in randomized controlled studies, and in 68.3–75 % of patients treated with valproate in non-randomized controlled trials.
Information on the safety of intravenous valproate derives from a limited number of dedicated safety studies, adverse event reporting in the efficacy studies described above, individual case reports, and pharmacovigilance reporting.
Dedicated Safety Studies
A number of dedicated studies have investigated the safety of intravenous valproate in patients with epilepsy during the clinical development of this formulation. Not all of these studies concern patients with SE, but the data obtained are nonetheless pertinent to the risk–benefit assessment of intravenous valproate in SE.
The largest of these studies included 318 patients, both children and adults, with epilepsy and hospitalized for seizure-related reasons . The initial dose was 15 mg/kg/day in four divided doses every 6 h. Subsequent doses were adjusted at the discretion of participating investigators as a function of serum valproate concentrations. The median dose of valproate administered was 375 mg infused over 1 h. The median number of doses was four, given over 2 days. Transient adverse events were reported in 54 patients (17.5 %), most frequently headache, injection-site reactions, nausea, somnolence, vomiting, dizziness, and taste perversion. None were reported in more than seven patients (2.3 %). No changes in hematological parameters, blood chemistry, or vital signs were observed.
A second large study compared infusion of valproate at either 1.5 or 3 mg/kg/min in 112 hospitalized children or adults with epilepsy . The maximum dose per infusion permitted was 15 mg/kg, and the maximum number of infusions per 24 h period was four. The primary outcome was changes in blood pressure during and after the first valproate infusion. Although no change in mean blood pressure was observed in the sample as a whole, two subjects infused at a rate of 3 mg/kg/min presented transient hypotension. The most common adverse events reported were somnolence, paresthesia, dizziness, and nausea. None of the patients had any alteration of consciousness. A possibly treatment-related case of encephalopathy was reported as a serious adverse event in a patient receiving valproate at an infusion rate of 3 mg/kg/min (peak serum valproate concentration of 123 µg/ml), which resolved upon discontinuation of valproate. No clinically significant abnormalities in blood chemistry or hematology were noted.
A number of other smaller safety studies [66–69] have reported adverse event profiles similar to these two large studies. In addition, one case of asymptomatic hyperammonemia has been described . More recently, a prospective study on 40 patients receiving an intravenous loading dose of 20 mg or 30 mg at 6 mg or 10 mg/kg/min of valproate found asymptomatic hyperammonemia in 30 of 40 patients 1 h after the infusion, decreasing to 66 % at 24 h after infusion . None of the patients had any alteration of consciousness or increase of transaminases.
Concerning more rapid injection schedules, one study  compared blood pressure changes and adverse events between 36 patients infused at a rate of 3 mg/kg/min up to a maximum dose of 15 mg/kg, 24 patients infused at 3 mg/kg/min up to 30 mg/kg, and six patients infused at 6 mg/kg/min up to 15 mg/kg. No increase in the incidence of adverse events was seen with the higher dose or faster infusion rate, and no changes in blood pressure were observed. Subsequently, infusion rates of up to 11 mg/kg/min have been reported in pediatric patients, with no untoward safety issue being identified .
Since hypotension is a concern with other intravenous AED treatments for SE, such as phenytoin, a retrospective analysis has been performed of the hemodynamic effects of intravenous valproate used to control SE in 13 patients with antecedents of hypotension or cardiovascular instability ; 12 of these patients were elderly (>64 years). The mean loading dose of intravenous valproate was 25.1 ± 5.0 mg/kg infused at a rate of 36.6 ± 25.1 mg/min, which produced serum valproate concentrations after completion of the infusion of 78.7 ± 35.8 mg/l. All except one patient received vasopressors (dopamine or dobutamine) before the valproate infusion began. No significant changes in systolic or diastolic blood pressure were observed during or after valproate infusion, and increases in vasopressor dose were not required. No other cardiovascular adverse effects or changes in cardiac rhythm were observed.
An open-label, prospective trial evaluated the safety of rapid intravenous loading of undiluted valproate (20 or 30 mg/kg/min at rates of 6 or 10 mg/kg/min) in 40 patients with epilepsy . Rapid administration of valproate was well tolerated, with no significant changes in heart rate and mean arterial pressure by dose group (20 mg/kg, 30 mg/kg) across time (measurements at 20, 30, 45, 60, and 240 min after administration). No patient exhibited alteration of consciousness at 30–60 min after dose administration compared with baseline, although three patients (7.5 %) complained of sedation and one of nausea. There were no disturbances in cardiac conduction or arrhythmias as indicated by normal electrocardiogram (ECG) in all patients at all time points. A total of 30 patients (81.5 %) complained of local irritation (pain/burning or paresthesias) lasting less than 3 min, with no indication of redness, irritation, or phlebitis. No significant changes in platelets or hepatic parameters were observed.
Although this study was not conducted in patients with SE, rapid administration of undiluted valproate proved safe and well tolerated, supporting its use in emergent situations.
The risk of local injection-site reactions with intravenous valproate was explored specifically in a retrospective chart review of two double-blind randomized clinical trials evaluating the use of valproate or phenytoin to prevent post-traumatic seizures in 775 patients with traumatic brain injury . Injection-site reactions occurred in 18 % of patients receiving valproate and 25 % of those receiving phenytoin, for the most part during administration of the initial loading dose. When only patients receiving the drugs by a peripheral line were taken into account, the frequency of injection-site reactions with valproate and phenytoin was 21 and 30 %, respectively (p = 0.056). No cases of skin necrosis or purple glove syndrome were observed in both groups. However, other studies reported an incidence of purple glove syndrome of 1.7–5.9 % after phenytoin administration [74, 75].
Comparative Randomized Studies
The six comparative randomized studies included 183 patients treated with valproate. The incidence of hypotension and respiratory depression was reported in all six studies, and the incidence of hepatic abnormalities in two studies. The overall rate of these three adverse events in patients treated with intravenous valproate was 0.5, 0.5, and 3.8 %, respectively (Table 5). The incidence of hypotension and respiratory depression was lower with valproate than with phenytoin (8.7 and 4.3 %, respectively) or diazepam (21.4 and 25 %, respectively). Conversely, the incidence of hypotension was equal to that of intravenous phenobarbital (0 %), whereas the incidence of respiratory depression was lower for valproate than that reported for intravenous phenobarbital (3.3 %). On the other hand, the incidence of liver abnormalities was higher than that reported for intravenous phenytoin (2.2 %). Given the low patient numbers, none of the differences reported in the individual studies were statistically significant.
A recent systematic review assessing the role of valproate in generalized convulsive SE included the studies comparing intravenous valproate with intravenous phenytoin [41, 42, 46] in a meta-analysis . Compared with phenytoin, intravenous valproate had a statistically lower risk of adverse effects (considered as a whole) (RR 0.31, 95 % CI 0.12–0.85).
Comparative Not-Randomized Studies
The four comparative not-randomized studies included 197 patients treated with valproate. The incidence of hypotension and respiratory depression was reported in only one study, showing no difference between the drugs (0 %) .
Uncontrolled Studies and Case Series
Adverse events reported in the other prospective studies and retrospective case series discussed above are presented in Table 6. Interpretation of these studies is limited by the disparate nature of adverse event reporting between studies and incomplete ascertainment. Nevertheless, the nature and frequency of the adverse events described are essentially similar to what was observed in the dedicated safety studies. The most frequently reported side effects were nausea/vomiting, dizziness, and sedation (Tables 7 and 8).
No effects on respiratory function were noted. Cardiovascular effects were limited to slight hypotension reported in five patients overall; this was generally mild and transient, and effects of co-medication were suspected in some cases. The local tolerability of intravenous valproate was also good, with only two subjects reporting injection-site pain. Mild hyperammonemia and mild thrombocytopenia were reported in four patients each.
A number of case reports have appeared in the literature describing idiosyncratic adverse reactions to intravenous valproate in patients with SE. Several of these have described cases of encephalopathy. For example, Rossetti and Bromfield  have described two patients developing acute encephalopathy within 48 h of receiving intravenous valproate for the treatment of SE. Serum valproate concentrations were in the therapeutic range. In these cases, encephalopathy was associated with hyperammonemia twofold higher than normal. Symptoms resolved after reduction of the valproate dose. Embacher et al.  described a woman who developed symptoms of encephalopathy within 24 h of receiving intravenous valproate. Again, serum valproate levels were not markedly elevated (29 µg/ml), and serum ammonium was within the normal range. The patient recovered completely within 48 h of discontinuation of valproate. Reversible symptoms of encephalopathy, including confusion or lethargy, were observed in three adult patients with absence SE treated with intravenous valproate and who were on maintenance therapy with lamotrigine . In these patients, the authors related the symptoms to elevated serum levels of lamotrigine due to the well characterized pharmacokinetic interaction between valproate and lamotrigine, whereby valproate inhibits hepatic glucuronidation of lamotrigine.
In addition to these reports of encephalopathy, a case of significant hypotension following intravenous valproate infusion was described in an 11-year-old girl presenting febrile SE following varicella infection and receiving valproate at a dose of 30 mg/kg over 1 h . This patient had previously received 10 mg of diazepam and then 1.6 mg of lorazepam, shortly before infusion of valproate was initiated in an attempt to control recurrent seizure activity. The contribution of valproate to the onset of hypotension is consequently difficult to determine.
Intravenous administration of valproate can be associated with acute pancreatitis  and with hepatic failure. A case of acute pancreatitis following intravenous valproate administration has been reported, as well as a case of fulminant hepatic failure associated with hemorrhagic shock leading to death in 9-year-old patient with cerebellar syndrome and moderate mental retardation receiving intravenous valproate for the treatment of SE . Four months after starting valproate treatment (initially as polytherapy up to 600 mg/day and later as monotherapy up to 1,200 mg/day), the child had nausea and vomiting, with subsequent increasing sleepiness and reduced vigilance. Laboratory exams revealed severe liver failure. Post-mortal examinations showed a microvescicular hepatic steatosis with centro- and mediozonal necrosis, hemoperitoneum, retroperitoneal bleeding, and hemorrhage involving pancreas, adrenal glands, and small bowel. The authors did not rule out the presence of underlying metabolic disorders (organic acidemias, mitochondrial disorders, Alper’s disease).
As of 31 January 2006, a total of 517 medically confirmed adverse drug reactions in 224 patients receiving intravenous valproate had been reported in the worldwide Sanofi-Aventis post-marketing pharmacovigilance database since 1994 . Given the estimated exposure to intravenous valproate over the period (>1 million units prescribed per year worldwide), the reporting rate for adverse events was less than one case per 100,000 administrations. It should be noted that the indication for prescription was not necessarily SE in all these cases, and in many cases the indication was not reported. SE was explicitly reported as the reason for prescription in around one-quarter of these cases. Children, again in all indications, accounted for 21 % of the reported events. The most frequently reported adverse drug reactions (between ten and 50 reports each) were hyperammonemia, coma, encephalopathy, metabolic encephalopathy, thrombocytopenia, confusional state, pancreatitis, hepatic failure, somnolence, and EEG abnormalities. Reporting of these events did not necessarily imply that a causal relationship to treatment had been identified. No relevant differences in the distribution of adverse drug reactions were noted between children and adults or between patients known to have received treatment for SE and the others.