Robertson 2002.

Methods	RCT Setting: UK
Participants	40 randomised but 36 seen for immediate outcome assessment (see Notes), recruited from London hospital and community rehabilitation teams, had left visual neglect on cancellation or bisection tests (The following data describe the initial 40 participants: see Notes) Experimental: n = 19, control: n = 21 Mean age (SD): experimental 69.3 years (9), control 67 years (9.4) Sex (male/female): experimental 13/6, control 16/5 Onset of neglect (SD): experimental 152.8 days (142.4), control 152.1 days (117.9) Inclusion: right hemisphere stroke, aged under 80, right‐handed, no history of major psychiatric/disease/disability that would prevent participation or contaminate results
Interventions	LAT wearing (on the wrist/leg/shoulder) an active limb activation device during perceptual training. The device emitted an auditory tone if no left‐sided movement was made, versus perceptual training wearing an inactive (no tone) limb activation device Both groups received training at their residence (usually own home) for 12 weeks for approximately 45 minutes per week The perceptual training for both groups involved working on visuoperceptual puzzles and reading tasks which implicitly but not explicitly involved advice to scan to the left For analysis of bottom‐up and top‐down rehabilitation approaches this review coded the experimental condition as bottom‐up
Outcomes	The study collected 3 types of outcomes: BI/Nottingham EADL Bergego rating scale of neglect Motricity index (total left body side) at 4 time points: immediately post‐training, 3 months, 6 months, 18 to 24 months. In addition the BIT, Comb and Razor personal neglect test, and the modified Landmark test were given at the first 3 time points For comparability with other studies this review used only the following outcome/time points: BI immediate and 6 months
Notes	Attrition: 36/40 followed up immediately (experimental 17, control 19); 32 at 6 months, 26 at 18 to 24 months Groups appeared appropriately matched for demographic and clinical baseline variables No information on number per group at 6 months. Know 4 lost but not whether all were from a single group so assumed worst case and subtracted 4 per group, i.e. conservative sample estimate of 28 not 32
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	Robertson 2002 confirmed that the recruiters were unaware of and unable to predict allocation concealment. Authors confirmed randomisation but did not specify the method used. Concealment was highly likely to have been achieved, although it could not be guaranteed
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinded outcome assessor.
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 (2 from each group) dropped out during treatment. 36 participants were followed up at 3 months, and 32 at 6 months. Of the 4 who dropped out at 6 month follow‐up, 2 had a further CVA, 1 died, and 1 refused. Low risk, loss to follow‐up unlikely to affect outcomes
Free of systematic differences in baseline characteristics of groups compared?	Low risk	Groups comparable on baseline characteristics. A slight difference on verbal memory but unlikely to be relevant to outcomes
Did authors adjust for baseline differences in their analyses?	Low risk	Not needed