Fanthome 1995.

Methods	RCT Setting: UK
Participants	18 (see Notes) RH stroke patients admitted to hospital Experimental n = 9, control n = 9 (The following data describe the 18 initial participants: see Notes) Mean age (SD): experimental 66.3 years (10.7), control 71.1 years (7.6) Sex (male/female): experimental 6/3, control 6/3 Time post‐onset (mean months): experimental 1.0, control 0.6 Inclusion: not blind; < 80 years of age; no history of dementia or psychiatric problems; not ill; right‐handedness; score > 6 on Abbreviated Mental Test; RH stroke; score < 130 on BIT
Interventions	4 weeks (2 hours 40 minutes per week) feedback of eye movements (wearing specially adapted glasses with auditory signal) versus 4 weeks no treatment For analysis of bottom‐up and top‐down rehabilitation approaches this review coded the experimental condition as top‐down
Outcomes	The study collected 3 types of outcomes: eye movements, conventional BIT subtests and behavioural BIT subtests, immediately post‐treatment (4 weeks) and 4 weeks later (8 weeks) For this review we used the 4‐week single letter cancellation test (for immediate outcomes), and the 8‐week BIT summary behavioural subtest scores (for persisting outcomes)
Notes	Personal communication supplied group data on BIT subtests for all but 1 control participant at 4 weeks (missing data, therefore n = 18 ‐ 1), and the information that assessor blinded to allocation. BIT behavioural data are for all 18 at 4 weeks but only 13 at 8 weeks. 8 weeks = post‐start of treatment, i.e. is a 4‐week follow‐up post‐end of treatment Single letter cancellation data are for number cancelled, i.e. higher numbers indicate better outcome Experimental and control groups appeared adequately matched on demographic and clinical data although control group slightly older than experimental, no baseline BIT data
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	High risk	Sealed opaque envelopes prepared from random number tables. Concleament of allocation can not be guaranteed as it was not done by a third party. The combination of a small sample size with no external randomisation meant that there was a potential risk to concealment Random number tables
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinded outcome assessor
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	1 participant was recruited but not included as "he could not move his eyes to the fixation points".  1 participant from the control group was excluded as he was discharged home outside the area of the hospital
Free of systematic differences in baseline characteristics of groups compared?	Low risk	Groups appear similar at baseline, and no significant differences were found
Did authors adjust for baseline differences in their analyses?	Low risk	N/A – no baseline differences