Helde 2005.

Methods	Randomised controlled trial (2 years follow‐up with general satisfaction measured 3 months after this)
Participants	114 adult patients attending a neurological clinic in Trondheim, Norway Mean age of participants was 35 and 40 years in intervention and control arms respectively; 42% were male
Interventions	Intervention: group education programme plus follow‐up teaching and support from an epilepsy nurse Control: "conventional treatment according to individual needs"
Outcomes	Primary outcome Quality of life, using QOLIE‐89 inventory at 24 months Secondary outcomes General patient satisfaction measured by Visual Analogue Scale (VAS) at 27 months Correlation between the reported general satisfaction (at 27 months) and change in QOLIE‐89 data (between baseline and 24 months)
Funding	Study supported by a grant from Glaxo‐SmithKline, Norway
Notes	—
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated block randomisation
Allocation concealment (selection bias)	Unclear risk	Randomisation was coordinated by a research centre, but the authors gave no further details of how the trial conducted randomisation, what blocks it used, or how it concealed allocation
Blinding (performance bias and detection bias) All outcomes	Low risk	Neither clinicians nor participants were blinded. However, interviews were conducted (and presumably analysed) by independent research assistants blinded to treatment allocation
Incomplete outcome data (attrition bias) All outcomes	Low risk	97% of participants were included in the intention‐to‐treat analysis
Selective reporting (reporting bias)	Low risk	Outcomes were stated to be outcomes derived from the QOLIE‐89 questionnaire. All scores are reported
Other bias	High risk	Small study with only 28 randomised participants. Power calculations and required sample size were not reported. Some differences in baseline characteristics are noted (proportion living alone and receiving one antiepileptic drug)
Overall risk of bias	Low risk	Computer generated block randomisation, no blinding, and relatively low levels of dropout with most participants included in the intention‐to‐treat analysis