Results of the search

The search strategies used revealed a total of 349 records for potential inclusion. One hundred and eleven duplicate records were removed. An additional 225 records were excluded due to their irrelevance. The full‐text articles were obtained for the remaining 13 records to assess their eligibility for inclusion. We excluded another nine full‐text articles for various reasons, as listed in Figure 1. Four studies remained eligible for inclusion in the review and were consequently used to formulate a descriptive analysis.

Figure 1

---

---

Study flow diagram.

Included studies

The total number of participants included was 197. Details of the included studies are reported in the Characteristics of included studies table.

The study conducted by Allen 1983 was a double‐blind, randomised, placebo‐controlled, cross‐over trial. The treatment/control period lasted nine weeks with an eight‐week wash‐out period between arms. The clobazam dosage employed was 30 mg at night for all participants, equivalent to 0.5 mg/kg, presuming a mean participant weight of 60 kg. The other AEDs being taken during the study period were carbamazepine, phenytoin, phenobarbitone and sodium valproate. The study sample size was limited with only 26 participants, of mean age 34 (range 18 to 60). The study included both participants with generalised‐ and focal‐onset seizures, although the exact numbers of participants with each seizure type was not specified.

Keene 1990 was also a small, double‐blind, randomised, placebo‐controlled, cross‐over study, recruiting 21 participants, aged 2 to 19 years (mean 11 years, number over 16 years undisclosed). At baseline, thirteen participants had generalised‐onset seizures and eight had focal‐onset seizures. Following an initial one‐month baseline period, participants were assigned to a sequence of placebo‐clobazam or clobazam‐placebo. Treatment periods were of three months with a one‐month wash‐out period. The initial dosage of clobazam was calculated as 0.5 mg/kg on commencement, increased to 1 mg/kg if no response was achieved. Furthermore, were the participants to develop excessive drowsiness, the dosage was decreased by 0.25 mg/kg/day. Although they do state that baseline serum AED levels were taken and doses adjusted to achieve therapeutic levels, there is no report as to which AEDs were being taken at baseline.

The only multi‐centre trial identified was conducted by Koeppen 1987 and took place across five European countries. This was also a randomised, placebo‐controlled, cross‐over study and was reported as being double‐blind. This was the largest of the studies identified, recruiting 129 participants, of which, 63 were randomised to clobazam then placebo and 66 to placebo then clobazam. There was a one‐month baseline period and each 'treatment' period was of three months duration with a one‐month wash‐out period. Clobazam doses ranged from 10 to 40 mg/day, equating to a dosage range of 0.17 to 0.67 mg/kg, presuming a mean participant weight of 60 kg. The baseline AEDs being taken were, in order of decreasing frequency: carbamazepine, phenobarbital, phenytoin, valproate and primidone. The baseline seizure types were 32 of generalised‐onset and 149 of focal‐onset, during the preceding year.

The fourth study, Schmidt 1986, was also a randomised, double‐blind, placebo‐controlled, cross‐over study. The sample size was limited, involving only 21 participants aged between 18 and 55 years (mean 38 years). The dosage of clobazam used was 10 mg three times a day, increasing to a maximum of 40 mg three times a day in 10 mg increments in the first month. The maximum dose was continued for months two and three. Clobazam was then decreased by 5 to 10 mg stages over the 4th month. The placebo was increased and decreased following the same guide, thus creating a gradual wash‐out period over one month. This resulted in a mean dose of 0.54 mg/kg (range 0.23 to 0.74 mg/kg). The study concluded with a one‐month observation period, during which both clobazam and placebo were removed. The baseline AEDs being taken were, in order of decreasing frequency: phenytoin, carbamazepine, primidone, phenobarbital, valproic acid, mephenytoin and clonazepam. The baseline seizure types were six of generalised onset and 29 of focal onset. It was not stated if this was during the one‐month observation period or another time‐frame.

Excluded studies

A total of 9 studies (Andrade 2009; Basu 2007; Conry 2009; NCT02134366; NCT02564952; NCT02565108; NCT02726919; Semah 2014; Vajda 1985) were excluded from the review. The reasons for exclusions are provided in the Characteristics of excluded studies table.

During the initial search in March 2007, one study (Vajda 1985) was excluded. Despite the study being a randomised, controlled trial with cross‐over design, investigating clobazam as an add‐on for drug‐resistant epilepsy, the study was a pilot study with only a three‐day treatment period, followed by a longer‐term cross‐over study of only four weeks, with no wash‐out period. It therefore, did not satisfy the inclusion criteria that specified that studies must have a minimum treatment period of eight weeks duration. Furthermore, only nine participants were randomised, of which, four were later transferred to an open trial. Only preliminary results were presented and these incorporated open, non‐randomised participants. No standard follow‐up period or outcome measures were described. Additionally, for the five participants in the randomised phase, no data were presented to determine seizure frequency or 50% reduction. The full study report was not published.

An updated search in November 2012 identified three additional studies (Andrade 2009; Basu 2007; Conry 2009) for potential inclusion.The studies were not, however, applicable for inclusion because: one study investigated as clobazam as a monotherapy (Andrade 2009), one was open‐label and therefore did not suffice the blinding inclusion criteria (Basu 2007), and one had a treatment period shorter than 8‐weeks in duration (Conry 2009)

An updated search in 2016 identified one further potentially eligible study (Semah 2014). This study was not applicable for inclusion because it was an open‐label study and thus lacked any method of blinding.

During the latest update, the four studies, previously classified as ongoing studies (NCT02134366; NCT02564952; NCT02565108; NCT02726919), were reclassified as excluded studies. Notably, two of the studies (NCT02564952; NCT02726919) were identified to be open‐label extension studies with single‐group assignment, thus making the studies ineligible for inclusion in the review. Another study (NCT02134366) had been terminated due to recruitment issues, resulting from a limited target population. Consequently, there were no results available for inclusion in the review. For the remaining study (NCT02565108), we noted that both treatment groups received clobazam. The intervention treatment group received clobazam in combination with cannabidiol, whilst the control group received clobazam with add‐on placebo. The study, therefore, did not satisfy the inclusion criteria.

Allocation

Three of the included studies failed to adequately describe both the method used to generate the randomisation sequence, as well as the method for allocation concealment. The multicentre study by Koeppen 1987 reported that the randomisation method was different for each centre, however, the specific methods for each centre were not disclosed. All three studies were therefore awarded an unclear risk of bias for both parameters, random sequence generation and allocation concealment.In contrast, Keene 1990 did state that the hospital pharmacy was responsible for treatment allocation whilst a blinded pharmacist was responsible for dispensing the study drug and was thus awarded a low risk of selection bias. Keene 1990 did not, however, provide any details on how the randomisation sequence was generated and, consequently, we judged that Keene 1990 remained at unclear risk of bias for this domain.

Blinding

Two studies, Allen 1983 and Schmidt 1986, effectively described their methods of blinding. Matching placebo was given to participants to ensure the blinding of both participants and study personnel. In both studies, participants self‐reported their seizure frequency and adverse events, and therefore, acted as the outcome assessors. As a result, the outcome assessors were, likewise, effectively blinded. Furthermore, study personnel, including those responsible for data entry and data analysis would, likewise, have been adequately blinded throughout their study involvement. The risk of bias for Allen 1983 and Schmidt 1986 was thus rated as low for both performance and detection bias. Conversely, Keene 1990 and Koeppen 1987 did not provide details of how the blinding of participants and study personnel was achieved. Both studies consequently received an unclear risk of bias rating for performance and detection bias.

Incomplete outcome data

The risk of bias associated with attrition varied between the four studies. The study by Allen 1983 was judged to have a low risk of bias. Attrition was declared as the study reported that six participants withdrew during the trial. Additional analysis was carried out whereby all participants who withdrew from the study were included in the statistical analyses and were classified as non‐responders,thus, intention‐to‐treat analysis had been performed. Similarly, Keene 1990 reported that two participants withdrew from the study. The results reported, however, included the total number of randomised participants, thus implying that intention‐to‐treat analysis had been conducted. Keene 1990 was likewise awarded a low risk of bias, with respect to attrition. In contrast, whilst Koeppen 1987 revealed that 21 participants withdrew from the study and another two participants provided incomplete outcome data, thereby reporting the attrition rate, Koeppen 1987 did not, however, compute efficacy data from these participants and, therefore, did not account for them in the analysis. As a result, intention‐to‐treat analysis was not performed and the study was deemed to have a unclear risk of attrition bias. Similarly, Schmidt 1986 reported that one participant was excluded from the study following randomisation, due to non‐compliance regarding seizure count and tablet count. No information was provided as to which treatment group the participant belonged to when they were withdrawn, and moreover, the participant was excluded from any efficacy analysis. Consequently, intention‐to‐treat analysis was not conducted. Schmidt 1986 was, likewise, awarded an unclear risk of bias rating for attrition bias.

Selective reporting

All four studies (Allen 1983; Keene 1990; Koeppen 1987; Schmidt 1986) failed to supply a trial protocol for the 'Risk of bias' assessment. The four studies were judged to have unclear risk of bias with respect to selective reporting. Three of the studies, Allen 1983; Koeppen 1987 and Schmidt 1986, neglected to define any specific efficacy outcomes within the methods sections of their texts. It was therefore unclear whether all intended outcomes had been measured and reported within the respective results sections. In contrast, the remaining study, by Keene 1990, specified in the methods section of their article that their efficacy outcome of interest was a 50% or greater reduction in seizure frequency. The results for this outcome were then fully reported within the results section of the text. We, however, suspect that this was not the only outcome intended to be measured and would expect that the study would also have at least recorded adverse events. We thus deemed it likely that there was selective reporting in this study and similarly awarded this study an unclear risk of bias for reporting bias.

Other potential sources of bias

In the study by Allen 1983, participants were assessed for baseline seizure rates for an undisclosed period of time and were defined as being 'drug‐resistant' although there was no description of how the diagnosis of epilepsy had been made. Participants were thus included if they were having four or more uncontrolled seizures a month. Moreover, there was no information provided as to the exact number of participants with each baseline seizure type, these being generalised‐onset and focal‐onset seizures, which were later described within the results taken from the two treatment periods. No exclusion criteria were presented. Additionally, there was no description of any adjustment of the dosage. Notably, however, whilst the mean serum clobazam concentration, taken during the last three weeks of treatment, was 0.33 µmol/L, there was a wide range (0.13 µmol/L to 0.99 µmol/L). This could indicate either: a wide range of participant‐dependent pharmacokinetic variability, variation in compliance, or undisclosed dosage alteration. As a result, Allen 1983 was assessed to have a unclear risk of bias with regards to other sources of bias.

By contrast, Keene 1990 described well the inclusion criteria (requiring more than four seizures per month) and exclusion criteria (no history of degenerative central nervous system disorder, brain tumour or poor compliance) for the study. The period for which participants were required to have suffered more than four seizures per month, and the number of AEDs failing to control seizures, however, were not defined. Furthermore, there is no description of how the diagnosis of epilepsy was made or the medical co‐morbidities, although the majority of included participants had some degree of mental retardation (17/21). Whilst the dosage adjustment regimen was comprehensively planned, no data were presented as to the proportion of participants who required dosage amendment in this way. Keene 1990, likewise, was deemed to have an unclear risk of bias originating from other sources.

For the Koeppen 1987 study, doses ranged between 10 mg/day and 40 mg/day clobazam, however, there was no clear protocol for dosage adjustment or standardisation of starting dose described. In contrast to the other studies, Koeppen 1987 did specify that epilepsy was diagnosed according to the International League Against Epilepsy guidelines although the inclusion criteria that participants must have at least three focal seizures a month was again for an undisclosed period. There was no report of the methods of recruitment, exclusion criteria or medical co‐morbidities. This trial is particularly poorly standardised between the centres with regard to recruitment, exclusion criteria, randomisation and clobazam dosages. Collectively, this led to us similarly awarding Koeppen 1987 an unclear risk of bias from other sources.

By comparison to the other studies, Schmidt 1986 did utilise strict inclusion criteria. Participants had to have been attending the clinic for at least seven years and have been diagnosed with focal epilepsy, according to the International Classification of Epilepsy. Participants were defined as drug‐resistant if they were still suffering with at least three seizures per month in the year preceding the trial, despite maximally tolerated doses of AED(s). Participants with progressive brain lesions and those with impaired capacity, limiting their ability to comply with treatment and record seizures accurately, were excluded. No medical co‐morbidities were reported in the trial. Although the study described a slow up‐titration to a maximal dose, it did not mention any methods for any adjustment of dosage, dependent on patient tolerability, upon reaching the maximum dosage. The study was therefore, likewise, assessed to have an unclear risk of bias with respect to other sources of bias.

Fifty per cent or greater reduction in seizure frequency

Two of the included studies, involving a total of 47 participants, reported the outcome, 50% or greater reduction in seizure frequency (summary of findings Table for the main comparison),

Of the 26 participants in the study conducted by Allen 1983, 15 (57.7%) were reported as having had a 50% or greater reduction in seizure frequency when receiving clobazam. No information was reported regarding the number of participants who had experienced a 50% or greater reduction in seizure frequency after receiving placebo. The authors did, however, perform an intention‐to‐treat analysis and reported that a significantly greater proportion of participants had a 50% or greater reduction in seizure frequency when receiving clobazam, compared to placebo (P < 0.01). Confidence intervals (95%) were not provided and there were insufficient data in the report to allow the review authors to estimate the confidence intervals or calculate the number of participants achieving the outcome during the placebo treatment arm.

The authors of Keene 1990 reported that 11/21 participants (52.4%) achieved a 50% or greater seizure reduction compared to baseline during the clobazam treatment period compared to none during the placebo period. No P value or confidence intervals were presented. Insufficient data were provided to allow us to carry out any further analysis.

No data regarding 50% or greater reduction in seizures were described in Koeppen 1987. The Schmidt 1986 authors did not report data for 50% or greater reduction in seizure frequency, but they did report that eight participants (40%) achieved 75% or greater reduction in seizure frequency. However, these data were collected only during the last month of the three‐month treatment period, not across the entire treatment period. Furthermore, it was not made clear if comparisons were being made between treatment and pre‐treatment periods, or between the clobazam and placebo period. Again, no information was provided about the number of participants that specifically experienced a 50% or greater reduction in seizure frequency when receiving placebo.

The evidence implies that patients are much more likely to attain a 50% or greater reduction in seizure frequency when receiving clobazam than when receiving placebo. This finding is, however, based upon the assumption that the absence of data regarding the number of participants who achieved a 50% of greater reduction in seizure frequency for the placebo arm meant that no participants achieved this outcome whilst receiving placebo. Consequently, the efficacy of clobazam has likely been overestimated.

Seizure frequency

Notably, none of the included studies separately reported the proportion of participants with focal‐onset seizures, or the proportion of participants with generalised‐onset seizures, that attained a 50% or greater reduction in seizure frequency. In an attempt to address the question of whether clobazam is equally as efficacious in managing both focal and generalised epilepsy, we instead extracted the mean seizure frequency for the two treatment arms, from the three included studies (involving 176 participants) that reported these data.

Allen 1983 found that mean seizure frequency was significantly lower during the clobazam treatment arm compared to during the placebo treatment arm for both focal‐onset seizures (with or without secondary generalisation) and for all seizure types (clobazam versus placebo: 17.5 versus 32.4 (P = 0.02) and 17.0 versus 29.9 (P = 0.002), respectively). There was no significant difference in the mean seizure frequency between the two treatment arms for participants with generalised‐onset seizures (clobazam versus placebo: 16.7 versus 24.2). The data presented for mean seizure frequency failed to include standard deviations or standard errors, hence preventing us from undertaking any further analysis. The individual frequency of seizure types was not stated, although the mean number of seizures per participant whilst in the placebo phase demonstrated a higher frequency of focal‐onset seizures than generalised‐onset seizures (32.4 versus 24.2, respectively).

Keene 1990 did not report seizure frequency. Data allowing any further analysis by the review authors were not reported. Whilst there were more participants with generalised‐ rather than focal‐onset seizures (13 versus eight, respectively), on analysis of best responders, only those with focal seizures were significant responders (P < 0.05). Keene 1990 stated in the text that "Patients with partial seizures tended to respond better to clobazam than did those with generalized seizures, but numbers in each seizure group were small".

Koeppen 1987 presented the mean seizure frequency in graphical form, no numerical data were reported. The mean seizure reduction was reported as being near to the level of significance for participants with complex focal seizures (P = 0.06), the most prevalent seizure type for this study, however, no other numerical data were presented to support this. Additionally, the study did not report the mean seizure reduction for any of the other subtypes of seizures and, therefore, no comparison could be made.

The Schmidt 1986 authors reported that the mean number of all seizure types during the maintenance phase (months two to three of both treatment arms), when participants were maintained on the full dose of clobazam or placebo, was significantly lower during the clobazam treatment phase than the placebo phase (clobazam versus placebo: 10.4 versus 21.5, P < 0.001). An analysis of seizure subgroups found that mean number of focal‐onset seizures (with or without secondary generalisation) were, likewise, significantly reduced (clobazam versus placebo: 10.4 versus 20.3, P < 0.01). The results for the generalised‐onset seizure group were not significant, although a 100% reduction in seizure frequency was noted (clobazam versus placebo: 0 versus 1.1). During the whole study period, the number of participants suffering focal‐onset seizures was greater than the number suffering generalised‐onset seizures (20 versus five, respectively).

Interestingly, two of the included studies (Allen 1983; Keene 1990) that independently assessed the efficacy of clobazam against the seizure subtypes separately reported that clobazam significantly reduced the frequency of focal‐onset seizures and was ineffective at significantly reducing the frequency of generalised‐onset seizures. Likewise, the third study (Schmidt 1986) stated that clobazam did not significantly reduce the frequency of generalised‐onset seizures, but, nevertheless, simultaneously reported that clobazam produced a 100% reduction in the mean number of generalised‐onset seizures. Notably, in both studies by Allen 1983 and Schmidt 1986, a lower frequency of generalised seizures during the placebo phase, compared to focal‐onset seizures, was recognised.

Overall, the general consensus appears to be that clobazam is more efficacious at reducing focal‐ rather than generalised‐onset seizures, however, the slightly contradictory reports challenge and generate uncertainty about this finding. The limited amount of data provided by the studies prevented us, the review authors, from being able to either repeat their statistical analysis or conduct our own meta‐analysis. If possible, this would have enabled us to reach a more conclusive finding.

Seizure freedom

Three studies, involving 175 participants, reported the outcome, seizure freedom (summary of findings Table for the main comparison).

In the study conducted by Allen 1983, three participants (11.5%) were reported to be seizure‐free during the clobazam treatment arm. No information was provided regarding the number of participants who were seizure‐free whilst receiving placebo, and consequently, again, we can only assume that no participants were.

In contrast, Koeppen 1987 reported that 20 out of the 129 (15.5%) participants achieved seizure freedom during the clobazam arm of the experiment and clarified that no participants were seizure‐free during the placebo arm.

Schmidt 1986 revealed that four (20%) participants were seizure‐free during treatment with clobazam. However, these data were collected during only the last month of the three‐month treatment period, not across the entire treatment period. Again, no information was provided about the number of participants that experienced seizure freedom when receiving placebo.

Collectively, the data extracted suggest that participants are more likely to attain seizure‐freedom when receiving clobazam than when receiving placebo. Similarly, this finding is based upon the assumption that no participants achieved seizure‐freedom whilst receiving placebo, despite that this was not specifically declared in two of the studies. Again, it is likely that the efficacy of clobazam could be overestimated in this review.

Treatment withdrawal

All four studies, including 197 participants, reported treatment withdrawal due to any reason (summary of findings Table for the main comparison).

The authors of Allen 1983 reported that six participants withdrew from the study. One participant withdrew during treatment with placebo for undocumented reasons. For the remaining five participants, it is assumed that they withdrew during the clobazam treatment arm, however, the timing of the withdrawal and the reasons for withdrawal are not provided.

The withdrawal data reported by Keene 1990 states that two participants withdrew during the clobazam phase due to "severe behavioural change which did not respond to lowering the drug dosage", however, no details of these behavioural changes are documented.

In Koeppen 1987, 21 participants withdrew from treatment, of which three left the hospital and two withdrew for reasons not specified. Of the remaining 16 participants, three withdrew due to adverse events from clobazam, one due to adverse events on placebo, four due to insufficient efficacy of clobazam and eight due to insufficient efficacy of placebo. Overall, 10 participants withdrew from treatment during the clobazam arm and 11 withdrew during the placebo arm of the study.

Schmidt 1986 reported that one participant withdrew due to "noncompliance with seizure count and tablet count", however, the authors did not confirm which treatment arm this participant was primarily randomised to. As a consequence, the data are uninformative.

Overall, the evidence appears to suggest that there is only a slightly increased risk of treatment withdrawal for patients receiving clobazam, opposed to placebo.

Treatment withdrawal due to adverse events

Only two of the included studies, comprising of 150 participants, reported this outcome (summary of findings Table for the main comparison).

As described earlier, Keene 1990 reported that two participants withdrew from treatment during the clobazam arm due to "severe behavioural change". In the study by Koeppen 1987, three participants withdrew due to adverse events during treatment with clobazam, compared to one participant withdrawing due to adverse events during treatment with placebo.

The evidence is very limited due to the extremely low number of events reported, but does suggest that patients are more likely to withdraw from treatment, specifically as a result of adverse events, when receiving clobazam than when receiving placebo.

Adverse events

Limited adverse events data were presented by Allen 1983 and Keene 1990. The former study, Allen 1983, did not report any specific adverse events quantitatively, however, did state that "Adverse effects occurred more often during the clobazam period". Specifically within the study, six participants required dose reduction due to adverse events during the clobazam phase, compared to two participants during the placebo phase. It was not, however, made clear whether these participants were in addition to the six who formally withdrew from the study. The only adverse events alluded to were mood changes which included: irritability, depression, and disinhibition.

Keene 1990 only acknowledged adverse events whilst describing the withdrawal of two participants during the clobazam phase. These withdrawals were due to "severe behavioural change which did not respond to a lowering of the dose". No other data were presented regarding the nature or frequency of adverse events.

The authors of Koeppen 1987 report that 36% (38/106) of participants suffered adverse events whilst on clobazam compared to 12% (13/106) on placebo. Drowsiness (71% clobazam and 16% placebo) and dizziness (26% clobazam and 6% placebo) were the adverse events most commonly reported. Notably, only three participants withdrew from the study due to adverse events, whilst on clobazam, suggesting that the events experienced were likely mild to moderate in severity.

A large proportion of participants reporting adverse events was found by Schmidt 1986; 85% of participants complained of symptoms, mainly drowsiness (40% clobazam and 10% placebo), vertigo (35% clobazam and 5% placebo) and depression (25% clobazam and 5% placebo), whilst receiving clobazam. In contrast, 12 participants (60%) reported adverse events during the placebo treatment arm. It was not reported as to whether those suffering adverse events withdrew from the trial.

Collectively, the data presented suggest that adverse events are more frequently reported by participants when receiving clobazam compared to placebo. It is not clear what the most common adverse events were, or what the severity of these adverse events might be, however, drowsiness was the most reported adverse event across two of the studies.

Quality of life

Only the study conducted by Koeppen 1987, including 129 participants, described quality of life data. Data were assessed using the Clinical Global Impression (CGI) scale. Koeppen 1987 reported that 60% (64/106) of participants had achieved a therapeutic effect whilst on clobazam compared to 11% (12/106) on placebo. Furthermore, utilising three analogue scales they reported a significant improvement in mood state whilst in the clobazam phase (P < 0.05), although there was also a significant increase in drowsiness (P < 0.05). There was, however, no explanation of the methods used to collect and analyse these data and, notably, an intent‐to‐treat population was not used for this analysis.

Tolerance

The only study to attempt an assessment of tolerance was that conducted by Allen 1983 who, whilst reporting evidence of no tolerance, only attempted to assess this by comparison between the first and last four weeks of the treatment period, using a method that was not described.

In contrast, despite that neither Keene 1990 or Schmidt 1986 mentioned a formal assessment of tolerance, both studies reported that participants developed a tolerance to the antiepileptic effects of clobazam. Keene 1990 did not report the exact number of participants who developed tolerance but did state that it was "a significant number of patients", adding later that the tolerance reported, however, remained "much less frequent than that reported in the literature". Meanwhile, Schmidt 1986 specified that four of nine participants developed tolerance within the three‐month treatment period.

Due to the lack of specific details provided by the studies, with regards to the methodology and the exact number of participants concerned, it is difficult to reach an overall judgement on the likelihood of patients developing tolerance after receiving clobazam.