Methods

Controlled trial using a 'responder enriched' design

First 2 study periods adopted a non‐randomised before‐after design

Second portion of the trial adopted a randomised, placebo‐controlled, double‐blind, parallel design

Only the second portion of this ’responder enriched’ trial was included

Participants

Individuals who were responders when taking add‐on stiripentol during a pre‐randomisation baseline period were randomly assigned to continue add‐on stiripentol or to add‐on placebo. All participants who entered the preceding study were children with focal epilepsy. 32 participants were randomly assigned: 17 to add‐on stiripentol and 15 to add‐on placebo

Add‐on stiripentol group: 7 male, 10 female (total 17 participants); age: 8 ± 3 years (mean ± standard deviation)

Add‐on placebo group: 11 male, 4 female (total 15 participants); age: 10.4 ± 3.4 years

Inclusion criteria for baseline period

• Partial seizures

• Receiving carbamazepine as co‐medication, with a benzodiazepine (clobazam or clonazepam) and/or vigabatrin administered in association

• Receiving ≥ 400 mg/d of carbamazepine

Inclusion criteria for randomised, placebo‐controlled, double‐blind, trial: participants had to be responders (i.e. ≥ 50% decrease in seizure frequency during third month of open period vs baseline) to be eligible for randomisation

Exclusion criteria for baseline period: participants receiving other drugs and those whose parents were unable to comply regularly with drug delivery and daily seizure diary

Exclusion criteria during double‐blind period

• Increase in seizure frequency during double‐blind period compared with pre‐randomisation period; participant should drop out on the day that the number of seizures during the double‐blind period reached that of the baseline period (normalised to 30 days)

• Significant increase in seizure severity during double‐blind vs open period (seizures more prolonged or cyanotic, or secondarily generalised, or resulting in a fall or a postictal deficit)

• Status epilepticus during double‐blind period

Interventions

• Add‐on stiripentol vs add‐on placebo

• First study period was a 1‐month baseline with single‐blind add‐on placebo

• Second period was a 4‐month open phase with open add‐on stiripentol

  • First 2 study periods adopted a non‐randomised before‐after design

• At end of open phase, responders were randomly assigned to add‐on stiripentol or add‐on placebo for a 2‐month double‐blind period

  • At baseline, add‐on placebo was added to current dose of carbamazepine (dose 1), which had not been modified during baseline. During open period, 50 mg/kg/d of add‐on stiripentol replaced add‐on placebo from the first day, twice daily, whereas the carbamazepine dose was decreased by 50% (dose 2). After 1 month of the open period, if a few seizures persisted and tolerability was acceptable, add‐on stiripentol dose was increased for the next 3 months according to minimum plasma concentration which was measured at steady state 2 weeks earlier: up to 90 mg/kg/d if plasma concentration of add‐on stiripentol < 10 mg/L, and up to 75 mg/kg/d if 10 < plasma concentration of add‐on stiripentol < 15 mg/L, but no increase if plasma concentration of add‐on stiripentol > 15 mg/L. At randomisation, add‐on stiripentol or add‐on placebo was administered double‐blind at the same dose as was administered during the last 3 months of the open period

  • In the add‐on placebo group, add‐on stiripentol was withdrawn over 3 weeks, whereas carbamazepine dose was increased to dose 1 by progressive escalation each week. In the add‐on stiripentol group, doses of add‐on stiripentol and carbamazepine remained unchanged.

• Length of follow‐up for randomised double‐blind phase was 2 months

Outcomes

• Primary endpoint: number of participants meeting escape criteria during the double‐blind period (see 'Exclusion criteria during the double‐blind period' under the section 'Participants')

• Secondary endpoint: percentage change in seizure frequency during second month of the double‐blind period vs baseline

Notes

• Trial was conducted at a single centre (France)

• All participants were refractory to the usual antiepileptic drugs (including valproate, carbamazepine, benzodiazepines and phenytoin), as well as to vigabatrin as a new drug

• Presence of refractory epilepsy was not considered among inclusion criteria

• No definition of refractory epilepsy was provided

• Conflicts of interest or study sponsor was not reported

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Participants were randomly assigned by a computer‐generated list

Allocation concealment (selection bias)

Low risk

Central allocation (pharmacy‐controlled randomisation). "Each patient received tablets of both stiripentol and "placebo of stiripentol" and tablets of both carbamazepine and "placebo of carbamazepine". "Individual tablets were prepared by the pharmacist"

Blinding of participants and personnel (performance bias)
All outcomes

Low risk

Second part of the trial was defined as “double blind”. "Each patient received tablets of both stiripentol and "placebo of stiripentol" and tablets of both carbamazepine and "placebo of carbamazepine". "Individual tablets were prepared by the pharmacist"

Blinding of outcome assessment (detection bias)
All outcomes

Low risk

"Each patient received tablets of both stiripentol and "placebo of stiripentol" and tablets of both carbamazepine and "placebo of carbamazepine". "Individual tablets were prepared by the pharmacist"

Incomplete outcome data (attrition bias)
All outcomes

High risk

Numbers of dropouts from each group were reported, along with reasons for dropout. However, number of dropouts in both arms was high (add‐on stiripentol and add‐on placebo) (53.3% vs 35.3%)

Selective reporting (reporting bias)

Low risk

Study protocol is not available, but published reports include all expected outcomes

Other bias

High risk

Through its 'responder‐enriched' design, this study resulted in a primary efficacy evaluation of an enriched population of participants, as a result of random assignment only of those who responded to open‐label treatment (high risk of selection bias). High risk of carry‐over and withdrawal effects