Herzog 2012.

Study characteristics
Methods	Double‐blind, phase III, parallel‐group, randomised (2:1 ratio) controlled trial conducted at 15 hospitals in the USA.
Participants	Female participants, aged 13 to 45 years old, with intractable seizures despite trials of > 2 antiepileptic drugs at therapeutic levels, and monthly menses with intervals of 23 to 35 days. 294 participants with catamenial or non‐catamenial epilepsy were recruited (randomisation stratified by catamenial or non‐catamenial epilepsy); only 130 participants with catamenial epilepsy were relevant to this review. Mean age +/− SD (range): progesterone: 31.4 +/− 8.68 (11 to 45); placebo: 32.31 +/− 8.50 (14 to 45). Duration of epilepsy (years): mean +/− SD (range): progesterone: 18.11 +/− 10.24 (1 to 39); placebo: 18.60 +/− 10.76 (1 to 37).
Interventions	Treatment consisted of identical progesterone 200 mg or placebo lozenges, taken 3 times daily on days 14 to 28 of treatment cycles. 130 randomised (124 analysed): progesterone: 85 randomised/79 analysed; placebo: 45 randomised/45 analysed. 3 baseline menstrual cycles; 3 treatment menstrual cycles were analysed.
Outcomes	Primary outcome: Per cent of responders for all seizures combined during treatment as compared to baseline. Secondary outcomes: Per cent of women who showed 50% reduction in average daily seizure frequency for the most severe seizure type and individual seizure types (secondary generalised motor seizures, complex partial seizures, simple partial seizures). Per cent of women who became seizure‐free. Change in average daily seizure frequency for all seizures combined, the most severe seizure type, and individual seizure types.
Funding	This research was supported by a National Institutes of Health (NIH) research grant: NIH NINDS R01 39466.
Conflict of Interest	Full disclosures are available in the journal article and on the online version of the journal article.
Notes	Body mass index, seizure type, age at onset of epilepsy, epilepsy focus, laterality and basis of localisation at baseline were also reported. No significant differences between groups.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Blocked randomisation (block size of 6), conducted separately for women with and without catamenial epilepsy, was conducted by an unblinded statistician.
Allocation concealment (selection bias)	Unclear risk	No information reported.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Study was double‐blinded, and placebo lozenges used.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	“Unblinded throughout the study (2000‐2010) were the unblinded statistician, research pharmacist and study safety monitor.” ClinicalTrials.gov entry also confirms that outcome assessors were blinded.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Attrition rates and reasons reported; the paper states that an intention‐to‐treat approach was used, but it is unclear why only 79 women (out of 85 randomised to progesterone) were included in the responder analyses.
Selective reporting (reporting bias)	Low risk	All outcomes specified in the methods section and on ClinicalTrials.gov are well reported in the results of the paper. All expected outcomes reported.
Other bias	High risk	Underpowered ‐ expected recruitment compared to actual recruitment. The trial was stopped early due to futility analyses showing that the blinded conditional power of the comparison for the primary outcome for that stratum dropped below 50%. The trial is underpowered for the original hypothesis, but was stopped for the benefit of the participants, and the authors have identified a biological problem with the original design. “Failure of the trial to prove the principal hypothesis may relate to the design that attempted to treat 3 patterns of catamenial epilepsy which likely differ in pathophysiology with a single treatment regimen.”