Morrow 1990.

Methods	Controlled before‐and‐after trial (12 months follow‐up) although reported as randomised controlled trial (see 'Risk of bias' table for more details)
Participants	232 patients with epilepsy or suspected epilepsy and referred to further services by their primary care physician (GP) in Glamorgan, Wales Mean age of participants was 30 and 32 years in non‐randomised and randomised participants respectively; 40% were male
Interventions	Intervention: attendance at a Specialist Epilepsy Unit Control: attendance at a neurology clinic
Outcomes	Outpatient attendance at baseline, 3, 6 and 12 months Seizure control (from review of case notes) at baseline, 3, 6 and 12 months Number and type of antiepilepsy drugs ("during study period": specific time periods not reported) Adverse drug effects (patients complaining of symptoms related to antiepilepsy drugs) at baseline, 3, 6 and 12 months Plasma drug concentrations at baseline and 12 months Visits to GP at baseline, 3, 6 and 12 months Use of inpatient services at baseline, 3, 6 and 12 months Self reported provision of advice and counselling at baseline and 12 months Patient satisfaction at baseline and 12 months Psychosocial, social and occupational factors as measured by the Hospital Anxiety and Depression scale at baseline and 12 months Study author states that information was derived via interview or questionnaire at baseline or review of case notes (after 12 months)
Funding	No details about funding provided
Notes	—
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	This was described as a randomised controlled trial by the study author but was treated as a controlled before‐and‐after trial for the purposes of this review as only 78% of participants were successfully randomised, with both the referring physician and the consultant to whom referred having case‐by‐case veto over randomisation.
Allocation concealment (selection bias)	High risk	There was considerable variation in the size of the intervention (n = 130) and comparison (n = 102) arms. This occurred not only because clinicians had to agree with each referral, but also because they could withdraw participants from the trial at any time. The fact that the comparator arm (usual care) had many fewer participants as a result of the vetoes being exercised suggests a perceived bias against the comparator from those involved in allocating participants.
Blinding (performance bias and detection bias) All outcomes	High risk	Clinicians were not blinded and based on the problems with randomisation and allocation, it would appear they had a strong bias towards the intervention over comparator. Although some outcome measures were derived from medical records and therefore less prone to bias, overall it would appear there was a high risk of bias from the lack of blinding.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Participant dropout rates were not reported in the trial.
Selective reporting (reporting bias)	High risk	Although follow‐up measurements were made at 3, 6, and 12 months, it was not clear how these repeated measures were accounted for, if at all; they were reported as a single endpoint.
Other bias	High risk	Power calculations and the required sample size were not reported. There were significant differences at baseline between participants who were randomised and not randomised.
Overall risk of bias	High risk	Failed randomisation and allocation, a lack of blinding, significant levels of dropout, the reporting of outcomes was problematic, and there were significant differences at baseline between participants who were randomised and not randomised.